diff --git a/.github/workflows/weekly-model-update.yml b/.github/workflows/weekly-model-update.yml
index 76db5f1b..12b03d9e 100644
--- a/.github/workflows/weekly-model-update.yml
+++ b/.github/workflows/weekly-model-update.yml
@@ -53,11 +53,14 @@ jobs:
       - name: Validate generated JSON
         run: python3 -m json.tool llmfit-core/data/hf_models.json > /dev/null
 
+      - name: Guard architecture metadata merge
+        run: python3 scripts/test_preserve_catalog_metadata.py
+
       - name: Set up Rust
         uses: dtolnay/rust-toolchain@29eef336d9b2848a0b548edc03f92a220660cdb8 # stable
 
-      - name: Validate against JSON Schema
-        run: cargo test -p llmfit-core hf_models_match_schema
+      - name: Validate catalog schema and known hybrid KV
+        run: cargo test -p llmfit-core hf_models_
 
       - name: Check for changes
         id: changes
diff --git a/llmfit-core/data/hf_models.json b/llmfit-core/data/hf_models.json
index 0ffe12bd..b3f45c97 100644
--- a/llmfit-core/data/hf_models.json
+++ b/llmfit-core/data/hf_models.json
@@ -791,7 +791,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 1024,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -1028,7 +1028,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 3000,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -1130,7 +1130,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -1138,14 +1138,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "mit",
     "is_moe": true,
@@ -1292,7 +1292,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -1300,13 +1300,13 @@
     "hf_downloads": 37252,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 25600,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -1786,7 +1786,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 50260,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -1802,7 +1802,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -1810,13 +1810,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 4,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 4,
+    "head_dim": 32,
+    "hidden_size": 128,
+    "vocab_size": 1536,
+    "moe_intermediate_size": 384,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -2209,7 +2209,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -2217,13 +2217,13 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-01-05",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 3,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 8,
+    "head_dim": 16,
+    "hidden_size": 128,
+    "vocab_size": 8000,
+    "moe_intermediate_size": 512,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -2237,7 +2237,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -2245,13 +2245,13 @@
     "hf_downloads": 93,
     "hf_likes": 0,
     "release_date": "2023-01-06",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 3,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 8,
+    "head_dim": 16,
+    "hidden_size": 128,
+    "vocab_size": 8000,
+    "moe_intermediate_size": 512,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -2617,13 +2617,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 9,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 2,
+    "head_dim": 32,
+    "hidden_size": 128,
+    "vocab_size": 2048,
+    "moe_intermediate_size": 512,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -3096,7 +3096,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 4096,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -3140,7 +3140,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -3148,13 +3148,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 9,
+    "num_attention_heads": 6,
+    "num_key_value_heads": 2,
+    "head_dim": 32,
+    "hidden_size": 192,
+    "vocab_size": 4096,
+    "moe_intermediate_size": 672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -3169,7 +3169,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -3177,13 +3177,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 9,
+    "num_attention_heads": 6,
+    "num_key_value_heads": 2,
+    "head_dim": 32,
+    "hidden_size": 192,
+    "vocab_size": 4096,
+    "moe_intermediate_size": 672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -3674,7 +3674,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 8000,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -3690,7 +3690,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -3698,12 +3698,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 9,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 8,
+    "head_dim": 32,
+    "hidden_size": 256,
+    "vocab_size": 4096,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -3982,7 +3982,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -3990,13 +3990,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 4,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 26,
+    "hidden_size": 312,
+    "vocab_size": 21128,
+    "moe_intermediate_size": 1200,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -4376,13 +4376,13 @@
     "hf_downloads": 12,
     "hf_likes": 2,
     "release_date": "2024-04-18",
-    "num_hidden_layers": null,
+    "num_hidden_layers": 8,
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "hidden_size": 512,
+    "vocab_size": 9,
+    "moe_intermediate_size": 1024,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -4742,13 +4742,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 8,
+    "num_attention_heads": 6,
+    "num_key_value_heads": 2,
+    "head_dim": 64,
+    "hidden_size": 384,
+    "vocab_size": 8192,
+    "moe_intermediate_size": 1024,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -5073,12 +5073,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 5,
+    "num_key_value_heads": 1,
+    "head_dim": 64,
+    "hidden_size": 320,
+    "vocab_size": 8202,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -5152,7 +5152,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -5160,13 +5160,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -5424,7 +5424,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 6000,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -5530,7 +5530,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -5538,13 +5538,13 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-19",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 32,
+    "hidden_size": 384,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -5586,7 +5586,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -5594,13 +5594,13 @@
     "hf_downloads": 81,
     "hf_likes": 0,
     "release_date": "2022-11-04",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 32,
+    "hidden_size": 384,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -6375,12 +6375,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 8,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 32,
+    "hidden_size": 256,
+    "vocab_size": 8192,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -6521,7 +6521,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "languages": [
@@ -6532,13 +6532,13 @@
     "hf_downloads": 5,
     "hf_likes": 2,
     "release_date": "2024-07-31",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 24,
+    "hidden_size": 384,
+    "vocab_size": 658,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -6553,7 +6553,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "languages": [
@@ -6564,13 +6564,13 @@
     "hf_downloads": 8,
     "hf_likes": 3,
     "release_date": "2024-07-31",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 24,
+    "hidden_size": 384,
+    "vocab_size": 668,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -6771,13 +6771,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 6,
+    "num_key_value_heads": 2,
+    "head_dim": 64,
+    "hidden_size": 384,
+    "vocab_size": 8192,
+    "moe_intermediate_size": 1152,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -7300,7 +7300,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "Text embeddings for RAG",
     "capabilities": [],
     "languages": [
@@ -7311,13 +7311,13 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-20",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 32,
+    "hidden_size": 384,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -7633,12 +7633,12 @@
     "hf_downloads": 99,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 20,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -7738,12 +7738,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 22,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -7773,12 +7773,12 @@
     "hf_downloads": 99,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 22,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -7843,12 +7843,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 23,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -7878,12 +7878,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 23,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -7948,12 +7948,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 24,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -7983,12 +7983,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 24,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8018,12 +8018,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 24,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8053,12 +8053,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 24,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8088,12 +8088,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 24,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8123,12 +8123,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 24,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8158,12 +8158,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 24,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8193,12 +8193,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 24,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8298,12 +8298,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 25,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8333,12 +8333,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 25,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8368,12 +8368,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 25,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8403,12 +8403,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 25,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8438,12 +8438,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 25,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8473,12 +8473,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 25,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8508,12 +8508,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 25,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8613,12 +8613,12 @@
     "hf_downloads": 94,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8648,12 +8648,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8683,12 +8683,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8718,12 +8718,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8753,12 +8753,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8788,12 +8788,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8823,12 +8823,12 @@
     "hf_downloads": 102,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8858,12 +8858,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8893,12 +8893,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8928,12 +8928,12 @@
     "hf_downloads": 99,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8963,12 +8963,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -8998,12 +8998,12 @@
     "hf_downloads": 99,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9033,12 +9033,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9068,12 +9068,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9103,12 +9103,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9138,12 +9138,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9173,12 +9173,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9208,12 +9208,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9278,12 +9278,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9313,12 +9313,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9348,12 +9348,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9383,12 +9383,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9418,12 +9418,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9453,12 +9453,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 26,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9523,12 +9523,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9558,12 +9558,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9593,12 +9593,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9628,12 +9628,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9663,12 +9663,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9698,12 +9698,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9733,12 +9733,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9768,12 +9768,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9803,12 +9803,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9838,12 +9838,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9873,12 +9873,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9908,12 +9908,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9943,12 +9943,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -9978,12 +9978,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10013,12 +10013,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10048,12 +10048,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10083,12 +10083,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10118,12 +10118,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10153,12 +10153,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10188,12 +10188,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10223,12 +10223,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10258,12 +10258,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10293,12 +10293,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10328,12 +10328,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10363,12 +10363,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10433,12 +10433,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10468,12 +10468,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10503,12 +10503,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10538,12 +10538,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10573,12 +10573,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10608,12 +10608,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10643,12 +10643,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10678,12 +10678,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10713,12 +10713,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10748,12 +10748,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10783,12 +10783,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 27,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10888,12 +10888,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10923,12 +10923,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -10958,12 +10958,12 @@
     "hf_downloads": 71,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11028,12 +11028,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11063,12 +11063,12 @@
     "hf_downloads": 92,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11098,12 +11098,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11133,12 +11133,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11168,12 +11168,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11203,12 +11203,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11238,12 +11238,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11273,12 +11273,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11308,12 +11308,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11343,12 +11343,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11378,12 +11378,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11413,12 +11413,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11448,12 +11448,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11483,12 +11483,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11518,12 +11518,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11553,12 +11553,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11623,12 +11623,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11658,12 +11658,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11693,12 +11693,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11728,12 +11728,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11763,12 +11763,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11798,12 +11798,12 @@
     "hf_downloads": 100,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11833,12 +11833,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11868,12 +11868,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11903,12 +11903,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11938,12 +11938,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -11973,12 +11973,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12008,12 +12008,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12043,12 +12043,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 28,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12218,12 +12218,12 @@
     "hf_downloads": 95,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12253,12 +12253,12 @@
     "hf_downloads": 99,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12288,12 +12288,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12323,12 +12323,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12358,12 +12358,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12393,12 +12393,12 @@
     "hf_downloads": 81,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12428,12 +12428,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12463,12 +12463,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12533,12 +12533,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12568,12 +12568,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12603,12 +12603,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12638,12 +12638,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12673,12 +12673,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12708,12 +12708,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12743,12 +12743,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12778,12 +12778,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12813,12 +12813,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12848,12 +12848,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12883,12 +12883,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12918,12 +12918,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12953,12 +12953,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -12988,12 +12988,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13023,12 +13023,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13058,12 +13058,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13093,12 +13093,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13128,12 +13128,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13163,12 +13163,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13198,12 +13198,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13233,12 +13233,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13268,12 +13268,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13303,12 +13303,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13338,12 +13338,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13373,12 +13373,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13408,12 +13408,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13443,12 +13443,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 29,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13548,12 +13548,12 @@
     "hf_downloads": 85,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13583,12 +13583,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13618,12 +13618,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13653,12 +13653,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13688,12 +13688,12 @@
     "hf_downloads": 101,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13723,12 +13723,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13758,12 +13758,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13793,12 +13793,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13825,12 +13825,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13860,12 +13860,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13895,12 +13895,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13930,12 +13930,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -13965,12 +13965,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14000,12 +14000,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14035,12 +14035,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14070,12 +14070,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14105,12 +14105,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14140,12 +14140,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14175,12 +14175,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14210,12 +14210,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14245,12 +14245,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14280,12 +14280,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14315,12 +14315,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14350,12 +14350,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14385,12 +14385,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14420,12 +14420,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14455,12 +14455,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14490,12 +14490,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14525,12 +14525,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14560,12 +14560,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14595,12 +14595,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14630,12 +14630,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14665,12 +14665,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14735,12 +14735,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14770,12 +14770,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14805,12 +14805,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14840,12 +14840,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14875,12 +14875,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -14910,12 +14910,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 30,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15015,12 +15015,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15050,12 +15050,12 @@
     "hf_downloads": 93,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15085,12 +15085,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15120,12 +15120,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15155,12 +15155,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15190,12 +15190,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15225,12 +15225,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15260,12 +15260,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15295,12 +15295,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15330,12 +15330,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15365,12 +15365,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15400,12 +15400,12 @@
     "hf_downloads": 99,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15435,12 +15435,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15470,12 +15470,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15505,12 +15505,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15540,12 +15540,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15575,12 +15575,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15610,12 +15610,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15645,12 +15645,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15680,12 +15680,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15715,12 +15715,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15750,12 +15750,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15785,12 +15785,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15820,12 +15820,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15855,12 +15855,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15890,12 +15890,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15925,12 +15925,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15960,12 +15960,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -15995,12 +15995,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16030,12 +16030,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16065,12 +16065,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16100,12 +16100,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 31,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16450,12 +16450,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16485,12 +16485,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16520,12 +16520,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16555,12 +16555,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16590,12 +16590,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16625,12 +16625,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16660,12 +16660,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16695,12 +16695,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16730,12 +16730,12 @@
     "hf_downloads": 104,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16762,12 +16762,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16797,12 +16797,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16832,12 +16832,12 @@
     "hf_downloads": 74,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16867,12 +16867,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16937,12 +16937,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -16972,12 +16972,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17007,12 +17007,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17042,12 +17042,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17077,12 +17077,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17112,12 +17112,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17147,12 +17147,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17182,12 +17182,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17217,12 +17217,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17252,12 +17252,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17287,12 +17287,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17322,12 +17322,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17357,12 +17357,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17392,12 +17392,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17427,12 +17427,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17462,12 +17462,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17497,12 +17497,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17532,12 +17532,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17567,12 +17567,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 32,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17777,12 +17777,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17812,12 +17812,12 @@
     "hf_downloads": 85,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17847,12 +17847,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17882,12 +17882,12 @@
     "hf_downloads": 86,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17917,12 +17917,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17952,12 +17952,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -17987,12 +17987,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18022,12 +18022,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18057,12 +18057,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18092,12 +18092,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18127,12 +18127,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18162,12 +18162,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18197,12 +18197,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18232,12 +18232,12 @@
     "hf_downloads": 93,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18302,12 +18302,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18337,12 +18337,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18372,12 +18372,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18404,12 +18404,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18439,12 +18439,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18474,12 +18474,12 @@
     "hf_downloads": 100,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18509,12 +18509,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18544,12 +18544,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18579,12 +18579,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18614,12 +18614,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18649,12 +18649,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18684,12 +18684,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18719,12 +18719,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18754,12 +18754,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18789,12 +18789,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18824,12 +18824,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18859,12 +18859,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18894,12 +18894,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18929,12 +18929,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18964,12 +18964,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -18999,12 +18999,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19034,12 +19034,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19069,12 +19069,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19104,12 +19104,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 33,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19209,12 +19209,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19244,12 +19244,12 @@
     "hf_downloads": 94,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19279,12 +19279,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19314,12 +19314,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19349,12 +19349,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19384,12 +19384,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19419,12 +19419,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19454,12 +19454,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19489,12 +19489,12 @@
     "hf_downloads": 97,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19524,12 +19524,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19559,12 +19559,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19594,12 +19594,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19629,12 +19629,12 @@
     "hf_downloads": 98,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19664,12 +19664,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19699,12 +19699,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19734,12 +19734,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19769,12 +19769,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19804,12 +19804,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19839,12 +19839,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19874,12 +19874,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19909,12 +19909,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19944,12 +19944,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -19979,12 +19979,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20014,12 +20014,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20049,12 +20049,12 @@
     "hf_downloads": 96,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20084,12 +20084,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20119,12 +20119,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20154,12 +20154,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20189,12 +20189,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20224,12 +20224,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20259,12 +20259,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20294,12 +20294,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20364,12 +20364,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20399,12 +20399,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20434,12 +20434,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20504,12 +20504,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20539,12 +20539,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20574,12 +20574,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 34,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20819,12 +20819,12 @@
     "hf_downloads": 93,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20854,12 +20854,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20889,12 +20889,12 @@
     "hf_downloads": 93,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20924,12 +20924,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20959,12 +20959,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -20994,12 +20994,12 @@
     "hf_downloads": 108,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21029,12 +21029,12 @@
     "hf_downloads": 88,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21064,12 +21064,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21099,12 +21099,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21169,12 +21169,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21204,12 +21204,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21239,12 +21239,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21274,12 +21274,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21309,12 +21309,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21344,12 +21344,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21379,12 +21379,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21414,12 +21414,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21449,12 +21449,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21484,12 +21484,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21519,12 +21519,12 @@
     "hf_downloads": 87,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21554,12 +21554,12 @@
     "hf_downloads": 99,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21589,12 +21589,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21624,12 +21624,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21659,12 +21659,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21694,12 +21694,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21729,12 +21729,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21764,12 +21764,12 @@
     "hf_downloads": 100,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21799,12 +21799,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21834,12 +21834,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21869,12 +21869,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21904,12 +21904,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21939,12 +21939,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -21974,12 +21974,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22009,12 +22009,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22044,12 +22044,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22079,12 +22079,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22114,12 +22114,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22149,12 +22149,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22184,12 +22184,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22219,12 +22219,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22254,12 +22254,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22289,12 +22289,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22324,12 +22324,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22359,12 +22359,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22391,12 +22391,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22426,12 +22426,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22461,12 +22461,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22493,12 +22493,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22528,12 +22528,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22563,12 +22563,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22598,12 +22598,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 35,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22703,12 +22703,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22738,12 +22738,12 @@
     "hf_downloads": 96,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22773,12 +22773,12 @@
     "hf_downloads": 94,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22808,12 +22808,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22843,12 +22843,12 @@
     "hf_downloads": 99,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22878,12 +22878,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22913,12 +22913,12 @@
     "hf_downloads": 108,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22948,12 +22948,12 @@
     "hf_downloads": 95,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -22983,12 +22983,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23015,12 +23015,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23050,12 +23050,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23120,12 +23120,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23155,12 +23155,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23190,12 +23190,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23225,12 +23225,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23260,12 +23260,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23295,12 +23295,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23330,12 +23330,12 @@
     "hf_downloads": 103,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23365,12 +23365,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23400,12 +23400,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23435,12 +23435,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23470,12 +23470,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23505,12 +23505,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23540,12 +23540,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23645,12 +23645,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23680,12 +23680,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23715,12 +23715,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23750,12 +23750,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23785,12 +23785,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23820,12 +23820,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23855,12 +23855,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23890,12 +23890,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23925,12 +23925,12 @@
     "hf_downloads": 93,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23960,12 +23960,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -23995,12 +23995,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24030,12 +24030,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24065,12 +24065,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24100,12 +24100,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24135,12 +24135,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24170,12 +24170,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24205,12 +24205,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24240,12 +24240,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24275,12 +24275,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24310,12 +24310,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24345,12 +24345,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 36,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24835,12 +24835,12 @@
     "hf_downloads": 84,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24870,12 +24870,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24905,12 +24905,12 @@
     "hf_downloads": 93,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24940,12 +24940,12 @@
     "hf_downloads": 93,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -24975,12 +24975,12 @@
     "hf_downloads": 86,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25010,12 +25010,12 @@
     "hf_downloads": 88,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25045,12 +25045,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25080,12 +25080,12 @@
     "hf_downloads": 95,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25115,12 +25115,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25150,12 +25150,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25220,12 +25220,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25255,12 +25255,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25290,12 +25290,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25325,12 +25325,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25360,12 +25360,12 @@
     "hf_downloads": 100,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25395,12 +25395,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25430,12 +25430,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25465,12 +25465,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25570,12 +25570,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25605,12 +25605,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25640,12 +25640,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25675,12 +25675,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25780,12 +25780,12 @@
     "hf_downloads": 100,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25815,12 +25815,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25850,12 +25850,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25885,12 +25885,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25920,12 +25920,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25955,12 +25955,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -25990,12 +25990,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26025,12 +26025,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26060,12 +26060,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26095,12 +26095,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26130,12 +26130,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26165,12 +26165,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26200,12 +26200,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26235,12 +26235,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26305,12 +26305,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26340,12 +26340,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26375,12 +26375,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26410,12 +26410,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26445,12 +26445,12 @@
     "hf_downloads": 99,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26480,12 +26480,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26515,12 +26515,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26550,12 +26550,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26585,12 +26585,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26620,12 +26620,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26655,12 +26655,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26690,12 +26690,12 @@
     "hf_downloads": 95,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26725,12 +26725,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26760,12 +26760,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26795,12 +26795,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26830,12 +26830,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26865,12 +26865,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26900,12 +26900,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -26935,12 +26935,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 37,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27424,12 +27424,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27459,12 +27459,12 @@
     "hf_downloads": 93,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27494,12 +27494,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27529,12 +27529,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27564,12 +27564,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27599,12 +27599,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27634,12 +27634,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27669,12 +27669,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27704,12 +27704,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27739,12 +27739,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27774,12 +27774,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27809,12 +27809,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27844,12 +27844,12 @@
     "hf_downloads": 99,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27879,12 +27879,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27914,12 +27914,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27949,12 +27949,12 @@
     "hf_downloads": 105,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -27984,12 +27984,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28019,12 +28019,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28054,12 +28054,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28089,12 +28089,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28124,12 +28124,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28159,12 +28159,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28194,12 +28194,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28229,12 +28229,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28264,12 +28264,12 @@
     "hf_downloads": 67,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28299,12 +28299,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28334,12 +28334,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28369,12 +28369,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28404,12 +28404,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28439,12 +28439,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28474,12 +28474,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28509,12 +28509,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28544,12 +28544,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 38,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28754,12 +28754,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28789,12 +28789,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28859,12 +28859,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28894,12 +28894,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28929,12 +28929,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28964,12 +28964,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -28999,12 +28999,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29034,12 +29034,12 @@
     "hf_downloads": 100,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29069,12 +29069,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29104,12 +29104,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29139,12 +29139,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29174,12 +29174,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29244,12 +29244,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29279,12 +29279,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29314,12 +29314,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29349,12 +29349,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29384,12 +29384,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29454,12 +29454,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29489,12 +29489,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29524,12 +29524,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29559,12 +29559,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29594,12 +29594,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29629,12 +29629,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29664,12 +29664,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29699,12 +29699,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29734,12 +29734,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29769,12 +29769,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29804,12 +29804,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -29839,12 +29839,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 39,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30111,12 +30111,12 @@
     "hf_downloads": 88,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30146,12 +30146,12 @@
     "hf_downloads": 96,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30181,12 +30181,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30216,12 +30216,12 @@
     "hf_downloads": 94,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30251,12 +30251,12 @@
     "hf_downloads": 93,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30286,12 +30286,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30321,12 +30321,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30356,12 +30356,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30426,12 +30426,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30461,12 +30461,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30496,12 +30496,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30531,12 +30531,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30566,12 +30566,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30636,12 +30636,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30671,12 +30671,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30706,12 +30706,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30741,12 +30741,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30776,12 +30776,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30811,12 +30811,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30846,12 +30846,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30881,12 +30881,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30916,12 +30916,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30951,12 +30951,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -30986,12 +30986,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31056,12 +31056,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31091,12 +31091,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31126,12 +31126,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31161,12 +31161,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 40,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31441,12 +31441,12 @@
     "hf_downloads": 85,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31476,12 +31476,12 @@
     "hf_downloads": 88,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31511,12 +31511,12 @@
     "hf_downloads": 94,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31546,12 +31546,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31581,12 +31581,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31616,12 +31616,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31651,12 +31651,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31721,12 +31721,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31756,12 +31756,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31791,12 +31791,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31826,12 +31826,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31861,12 +31861,12 @@
     "hf_downloads": 104,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31896,12 +31896,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31931,12 +31931,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -31966,12 +31966,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32001,12 +32001,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32036,12 +32036,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32071,12 +32071,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32106,12 +32106,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32141,12 +32141,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32176,12 +32176,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32211,12 +32211,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32246,12 +32246,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32281,12 +32281,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32316,12 +32316,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32351,12 +32351,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32386,12 +32386,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32421,12 +32421,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32456,12 +32456,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 41,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32561,12 +32561,12 @@
     "hf_downloads": 73,
     "hf_likes": 4,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32596,12 +32596,12 @@
     "hf_downloads": 108,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32631,12 +32631,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32701,12 +32701,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32806,12 +32806,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32841,12 +32841,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32876,12 +32876,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32911,12 +32911,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -32946,12 +32946,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33013,12 +33013,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33048,12 +33048,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33080,12 +33080,12 @@
     "hf_downloads": 100,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33115,12 +33115,12 @@
     "hf_downloads": 96,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33150,12 +33150,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33185,12 +33185,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33220,12 +33220,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33255,12 +33255,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33290,12 +33290,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33325,12 +33325,12 @@
     "hf_downloads": 94,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33360,12 +33360,12 @@
     "hf_downloads": 76,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33395,12 +33395,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33430,12 +33430,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33462,12 +33462,12 @@
     "hf_downloads": 63,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 42,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33742,12 +33742,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33847,12 +33847,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33882,12 +33882,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33917,12 +33917,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33952,12 +33952,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -33987,12 +33987,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34022,12 +34022,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34057,12 +34057,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34127,12 +34127,12 @@
     "hf_downloads": 100,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34162,12 +34162,12 @@
     "hf_downloads": 101,
     "hf_likes": 2,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34197,12 +34197,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34267,12 +34267,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34302,12 +34302,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34337,12 +34337,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34372,12 +34372,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34407,12 +34407,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34477,12 +34477,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 43,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34687,12 +34687,12 @@
     "hf_downloads": 93,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34722,12 +34722,12 @@
     "hf_downloads": 108,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34757,12 +34757,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34792,12 +34792,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34827,12 +34827,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34862,12 +34862,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34897,12 +34897,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -34967,12 +34967,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35002,12 +35002,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35072,12 +35072,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35107,12 +35107,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35142,12 +35142,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35177,12 +35177,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35212,12 +35212,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35244,12 +35244,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35279,12 +35279,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35314,12 +35314,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 44,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35559,12 +35559,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35594,12 +35594,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35629,12 +35629,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35664,12 +35664,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35696,12 +35696,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35731,12 +35731,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35766,12 +35766,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35801,12 +35801,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35836,12 +35836,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35871,12 +35871,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35906,12 +35906,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35941,12 +35941,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -35976,12 +35976,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36011,12 +36011,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36046,12 +36046,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36081,12 +36081,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36116,12 +36116,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36151,12 +36151,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36186,12 +36186,12 @@
     "hf_downloads": 74,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36221,12 +36221,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36256,12 +36256,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36291,12 +36291,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36326,12 +36326,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 45,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36501,12 +36501,12 @@
     "hf_downloads": 96,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36536,12 +36536,12 @@
     "hf_downloads": 93,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36571,12 +36571,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36606,12 +36606,12 @@
     "hf_downloads": 97,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36641,12 +36641,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36676,12 +36676,12 @@
     "hf_downloads": 95,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36711,12 +36711,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36746,12 +36746,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36781,12 +36781,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36816,12 +36816,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36851,12 +36851,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36886,12 +36886,12 @@
     "hf_downloads": 100,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36921,12 +36921,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36956,12 +36956,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -36991,12 +36991,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37026,12 +37026,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37061,12 +37061,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37096,12 +37096,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37131,12 +37131,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37166,12 +37166,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37201,12 +37201,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37236,12 +37236,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37271,12 +37271,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37306,12 +37306,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37338,12 +37338,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37373,12 +37373,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37408,12 +37408,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 46,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37513,12 +37513,12 @@
     "hf_downloads": 68,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37548,12 +37548,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37583,12 +37583,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37618,12 +37618,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37653,12 +37653,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37688,12 +37688,12 @@
     "hf_downloads": 101,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37723,12 +37723,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37758,12 +37758,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37828,12 +37828,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37863,12 +37863,12 @@
     "hf_downloads": 87,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37898,12 +37898,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37933,12 +37933,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -37965,12 +37965,12 @@
     "hf_downloads": 88,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38000,12 +38000,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38035,12 +38035,12 @@
     "hf_downloads": 63,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38070,12 +38070,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38105,12 +38105,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 47,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38350,12 +38350,12 @@
     "hf_downloads": 27,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38385,12 +38385,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38420,12 +38420,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38455,12 +38455,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38490,12 +38490,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38525,12 +38525,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38560,12 +38560,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38595,12 +38595,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38630,12 +38630,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38665,12 +38665,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38700,12 +38700,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38735,12 +38735,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38770,12 +38770,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38805,12 +38805,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38840,12 +38840,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -38875,12 +38875,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 48,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39050,12 +39050,12 @@
     "hf_downloads": 95,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 49,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39085,12 +39085,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 49,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39120,12 +39120,12 @@
     "hf_downloads": 99,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 49,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39155,12 +39155,12 @@
     "hf_downloads": 98,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 49,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39190,12 +39190,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 49,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39225,12 +39225,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 49,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39260,12 +39260,12 @@
     "hf_downloads": 71,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 49,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39295,12 +39295,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 49,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39330,12 +39330,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39365,12 +39365,12 @@
     "hf_downloads": 95,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39400,12 +39400,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39435,12 +39435,12 @@
     "hf_downloads": 97,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39470,12 +39470,12 @@
     "hf_downloads": 97,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39505,12 +39505,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39540,12 +39540,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39575,12 +39575,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39610,12 +39610,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39645,12 +39645,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39680,12 +39680,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39715,12 +39715,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39750,12 +39750,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39785,12 +39785,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 50,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39855,12 +39855,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39890,12 +39890,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39925,12 +39925,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39960,12 +39960,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -39995,12 +39995,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40030,12 +40030,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40065,12 +40065,12 @@
     "hf_downloads": 71,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40100,12 +40100,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40135,12 +40135,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40170,12 +40170,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40205,12 +40205,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40240,12 +40240,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 51,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40310,12 +40310,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 52,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40345,12 +40345,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 52,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40380,12 +40380,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 52,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40450,12 +40450,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 53,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40485,12 +40485,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 53,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40520,12 +40520,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 53,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40555,12 +40555,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 53,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40590,12 +40590,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 53,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40625,12 +40625,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 53,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40660,12 +40660,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 54,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40695,12 +40695,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 54,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40765,12 +40765,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 54,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40800,12 +40800,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 54,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40835,12 +40835,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 54,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40905,12 +40905,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 54,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40940,12 +40940,12 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 55,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -40975,12 +40975,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 55,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41010,12 +41010,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 55,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41045,12 +41045,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 55,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41080,12 +41080,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 55,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41115,12 +41115,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 56,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41150,12 +41150,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 56,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41185,12 +41185,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 56,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41255,12 +41255,12 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 57,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41290,12 +41290,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 57,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41325,12 +41325,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 57,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41360,12 +41360,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 57,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41395,12 +41395,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 57,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41430,12 +41430,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 57,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41465,12 +41465,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 57,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41570,12 +41570,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 58,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41605,12 +41605,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 58,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41640,12 +41640,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 58,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41710,12 +41710,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 59,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41745,12 +41745,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 59,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41780,12 +41780,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 59,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41815,12 +41815,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 59,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41850,12 +41850,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 59,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -41885,12 +41885,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 60,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42060,12 +42060,12 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 61,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42095,12 +42095,12 @@
     "hf_downloads": 108,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 61,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42130,12 +42130,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 61,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42165,12 +42165,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 61,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42200,12 +42200,12 @@
     "hf_downloads": 101,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 61,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42235,12 +42235,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 61,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42270,12 +42270,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 61,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42305,12 +42305,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 61,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42340,12 +42340,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 61,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42375,12 +42375,12 @@
     "hf_downloads": 104,
     "hf_likes": 2,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 62,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42410,12 +42410,12 @@
     "hf_downloads": 100,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 62,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42445,12 +42445,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 62,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42480,12 +42480,12 @@
     "hf_downloads": 87,
     "hf_likes": 2,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 63,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42515,12 +42515,12 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 63,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42550,12 +42550,12 @@
     "hf_downloads": 103,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 63,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42585,12 +42585,12 @@
     "hf_downloads": 102,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 63,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42620,12 +42620,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 63,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42655,12 +42655,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 63,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42725,12 +42725,12 @@
     "hf_downloads": 86,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 64,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42760,12 +42760,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 64,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42830,12 +42830,12 @@
     "hf_downloads": 85,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 64,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42865,12 +42865,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 64,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42900,12 +42900,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 65,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42935,12 +42935,12 @@
     "hf_downloads": 105,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 66,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -42970,12 +42970,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 66,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43005,12 +43005,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 66,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43075,12 +43075,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 67,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43107,12 +43107,12 @@
     "hf_downloads": 90,
     "hf_likes": 1,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 68,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43142,12 +43142,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 70,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43212,12 +43212,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 72,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43247,12 +43247,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 78,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43314,12 +43314,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 144,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43349,12 +43349,12 @@
     "hf_downloads": 104,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 153,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43381,12 +43381,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 156,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43416,12 +43416,12 @@
     "hf_downloads": 105,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 175,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43483,12 +43483,12 @@
     "hf_downloads": 106,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 178,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43518,12 +43518,12 @@
     "hf_downloads": 107,
     "hf_likes": 0,
     "release_date": "2023-09-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 2,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 192,
+    "vocab_size": 185,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -43638,7 +43638,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -43646,13 +43646,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 9,
+    "num_key_value_heads": 3,
+    "head_dim": 64,
+    "hidden_size": 576,
+    "vocab_size": 32770,
+    "moe_intermediate_size": 1728,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -44038,7 +44038,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -44049,13 +44049,13 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-10-10",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 2,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32005,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -44407,7 +44407,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -44415,13 +44415,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 14,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 2560,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -44523,7 +44523,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -44531,13 +44531,13 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 8,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 512,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -44638,7 +44638,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -44646,13 +44646,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 14,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 2560,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -44741,7 +44741,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 16000,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -44936,7 +44936,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -44944,13 +44944,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 64,
+    "hidden_size": 512,
+    "vocab_size": 8192,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -45172,7 +45172,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -45180,13 +45180,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 14,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 2560,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -45230,7 +45230,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -45238,13 +45238,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 13824,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -45259,7 +45259,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -45267,13 +45267,13 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2022-11-08",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 8,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 512,
+    "vocab_size": 50265,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-4.0"
@@ -45288,7 +45288,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -45296,13 +45296,13 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2022-11-08",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 8,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 512,
+    "vocab_size": 50265,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-4.0"
@@ -45375,7 +45375,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -45386,13 +45386,13 @@
     "hf_downloads": 84,
     "hf_likes": 0,
     "release_date": "2023-03-07",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 8,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 512,
+    "vocab_size": 50099,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-4.0"
@@ -45407,7 +45407,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -45418,13 +45418,13 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-05-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 8,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 512,
+    "vocab_size": 50099,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-4.0"
@@ -45500,7 +45500,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 5120,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -45508,13 +45508,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 512,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -45586,7 +45586,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -45594,13 +45594,13 @@
     "hf_downloads": 14138,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 512,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -45934,7 +45934,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -45945,13 +45945,13 @@
     "hf_downloads": 90,
     "hf_likes": 1,
     "release_date": "2023-10-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 4,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32005,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -46113,7 +46113,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -46121,13 +46121,13 @@
     "hf_downloads": 1009,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 6656,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -46199,7 +46199,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 3072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -46207,13 +46207,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 64,
+    "hidden_size": 512,
+    "vocab_size": 12288,
+    "moe_intermediate_size": 1920,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -46256,7 +46256,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -46264,13 +46264,13 @@
     "hf_downloads": 94,
     "hf_likes": 0,
     "release_date": "2023-09-04",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 32,
+    "hidden_size": 384,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -46284,7 +46284,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -46292,13 +46292,13 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-09-04",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 32,
+    "hidden_size": 384,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -46887,7 +46887,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -46895,13 +46895,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 14,
+    "num_attention_heads": 9,
+    "num_key_value_heads": 1,
+    "head_dim": 64,
+    "hidden_size": 576,
+    "vocab_size": 16388,
+    "moe_intermediate_size": 1664,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -47233,7 +47233,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -47253,13 +47253,13 @@
     "hf_downloads": 537,
     "hf_likes": 8,
     "release_date": "2026-06-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 14,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 2560,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -47431,7 +47431,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 99,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -47552,7 +47552,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 30522,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -47580,7 +47580,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 30522,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -47626,7 +47626,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -47634,13 +47634,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 6656,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -47684,7 +47684,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -47692,13 +47692,13 @@
     "hf_downloads": 85,
     "hf_likes": 2,
     "release_date": "2022-05-23",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -47906,7 +47906,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -47917,13 +47917,13 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-10-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32005,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -48756,7 +48756,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -48764,13 +48764,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 6656,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -49024,7 +49024,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -49035,13 +49035,13 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 8,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -49126,7 +49126,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 50257,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -49455,7 +49455,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -49466,13 +49466,13 @@
     "hf_downloads": 83,
     "hf_likes": 0,
     "release_date": "2023-10-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 8,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32005,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -50500,7 +50500,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 50257,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -50916,7 +50916,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -50924,13 +50924,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 64,
+    "hidden_size": 512,
+    "vocab_size": 32768,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "appvoid-noncommercial-model-license-1.0"
@@ -51161,7 +51161,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 50257,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -51262,7 +51262,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -51270,13 +51270,13 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 8002,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -51539,13 +51539,13 @@
     "hf_downloads": 84,
     "hf_likes": 1,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -51571,13 +51571,13 @@
     "hf_downloads": 84,
     "hf_likes": 0,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -51600,13 +51600,13 @@
     "hf_downloads": 60,
     "hf_likes": 4,
     "release_date": "2022-06-20",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -51974,7 +51974,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -51985,13 +51985,13 @@
     "hf_downloads": 88,
     "hf_likes": 0,
     "release_date": "2023-10-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 10,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32005,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -52043,13 +52043,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 16000,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -52297,7 +52297,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -52305,13 +52305,13 @@
     "hf_downloads": 1549,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 6656,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -52444,7 +52444,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -52452,13 +52452,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 10,
+    "num_attention_heads": 10,
+    "num_key_value_heads": 10,
+    "head_dim": 64,
+    "hidden_size": 640,
+    "vocab_size": 32013,
+    "moe_intermediate_size": 2240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -52646,7 +52646,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -52654,13 +52654,13 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2022-05-06",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 21128,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -52915,7 +52915,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -52923,13 +52923,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 24000,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -53298,7 +53298,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -53306,13 +53306,13 @@
     "hf_downloads": 88,
     "hf_likes": 0,
     "release_date": "2023-08-25",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 28996,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -53383,7 +53383,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -53394,13 +53394,13 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-05-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 28996,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "afl-3.0"
@@ -53852,7 +53852,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -53860,13 +53860,13 @@
     "hf_downloads": 86,
     "hf_likes": 4,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -53881,7 +53881,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -53889,13 +53889,13 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-05-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -54099,7 +54099,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -54107,13 +54107,13 @@
     "hf_downloads": 74,
     "hf_likes": 5,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -54127,7 +54127,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -54135,13 +54135,13 @@
     "hf_downloads": 82,
     "hf_likes": 0,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -54212,7 +54212,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -54223,13 +54223,13 @@
     "hf_downloads": 92,
     "hf_likes": 5,
     "release_date": "2023-04-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -54244,7 +54244,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -54255,13 +54255,13 @@
     "hf_downloads": 85,
     "hf_likes": 8,
     "release_date": "2023-06-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -54276,7 +54276,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -54284,13 +54284,13 @@
     "hf_downloads": 94,
     "hf_likes": 1,
     "release_date": "2023-07-14",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -54305,7 +54305,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "Text embeddings for RAG",
     "capabilities": [],
     "languages": [
@@ -54316,13 +54316,13 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-09-20",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -54422,7 +54422,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -54430,13 +54430,13 @@
     "hf_downloads": 88,
     "hf_likes": 0,
     "release_date": "2023-04-20",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30527,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -54484,7 +54484,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -54495,13 +54495,13 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-05-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "afl-3.0"
@@ -54516,7 +54516,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -54527,13 +54527,13 @@
     "hf_downloads": 91,
     "hf_likes": 0,
     "release_date": "2023-05-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "afl-3.0"
@@ -54669,7 +54669,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "languages": [
@@ -54680,13 +54680,13 @@
     "hf_downloads": 83,
     "hf_likes": 0,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -54842,7 +54842,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -54853,13 +54853,13 @@
     "hf_downloads": 59,
     "hf_likes": 0,
     "release_date": "2023-06-19",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 31002,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -54918,7 +54918,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 32128,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -54963,7 +54963,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -54971,13 +54971,13 @@
     "hf_downloads": 89,
     "hf_likes": 2,
     "release_date": "2022-03-16",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 31090,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -55177,7 +55177,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "Text embeddings for RAG",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -55185,13 +55185,13 @@
     "hf_downloads": 0,
     "hf_likes": 2,
     "release_date": "2026-08-16",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -55349,7 +55349,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -55357,13 +55357,13 @@
     "hf_downloads": 64,
     "hf_likes": 1,
     "release_date": "2022-05-31",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -55377,7 +55377,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -55385,13 +55385,13 @@
     "hf_downloads": 92,
     "hf_likes": 0,
     "release_date": "2023-03-08",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -55662,7 +55662,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -55670,13 +55670,13 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-04-04",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32102,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -55823,13 +55823,13 @@
     "hf_downloads": 85,
     "hf_likes": 0,
     "release_date": "2023-08-30",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 29794,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -56048,7 +56048,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -56060,13 +56060,13 @@
     "hf_downloads": 72,
     "hf_likes": 5,
     "release_date": "2023-02-07",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 36000,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -56110,7 +56110,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -56118,13 +56118,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 50280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -56195,7 +56195,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -56203,13 +56203,13 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2022-08-19",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 39001,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -56364,7 +56364,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -56372,13 +56372,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 39930,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -56420,7 +56420,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -56428,13 +56428,13 @@
     "hf_downloads": 53,
     "hf_likes": 1,
     "release_date": "2025-01-27",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 4096,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -56504,7 +56504,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -56515,13 +56515,13 @@
     "hf_downloads": 1476,
     "hf_likes": 5,
     "release_date": "2026-07-30",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 8,
+    "num_attention_heads": 6,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 768,
+    "vocab_size": 32768,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -56723,7 +56723,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 50257,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -56825,7 +56825,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -56833,13 +56833,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 49152,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -56854,7 +56854,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -56862,13 +56862,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 49152,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -57502,12 +57502,12 @@
     "hf_downloads": 88,
     "hf_likes": 0,
     "release_date": "2023-06-12",
-    "num_hidden_layers": null,
+    "num_hidden_layers": 2,
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 30000,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -57579,7 +57579,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -57590,13 +57590,13 @@
     "hf_downloads": 67,
     "hf_likes": 2,
     "release_date": "2023-05-29",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 50256,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -58432,7 +58432,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 50261,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -58466,7 +58466,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 50261,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -58964,7 +58964,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -58972,13 +58972,13 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 50265,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -58992,7 +58992,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -59000,13 +59000,13 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2023-04-03",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 50265,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -59076,7 +59076,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -59084,13 +59084,13 @@
     "hf_downloads": 88,
     "hf_likes": 1,
     "release_date": "2022-05-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 50265,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-4.0"
@@ -59105,7 +59105,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -59113,13 +59113,13 @@
     "hf_downloads": 89,
     "hf_likes": 0,
     "release_date": "2022-05-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 50265,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-4.0"
@@ -59134,7 +59134,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -59142,13 +59142,13 @@
     "hf_downloads": 87,
     "hf_likes": 0,
     "release_date": "2022-05-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 50265,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-4.0"
@@ -59163,7 +59163,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -59171,13 +59171,13 @@
     "hf_downloads": 85,
     "hf_likes": 0,
     "release_date": "2022-05-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 50265,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-4.0"
@@ -59192,7 +59192,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -59204,13 +59204,13 @@
     "hf_downloads": 85,
     "hf_likes": 4,
     "release_date": "2023-02-07",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 50265,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -59225,7 +59225,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -59236,13 +59236,13 @@
     "hf_downloads": 84,
     "hf_likes": 4,
     "release_date": "2023-03-04",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 50265,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -59592,12 +59592,12 @@
     "hf_downloads": 89,
     "hf_likes": 1,
     "release_date": "2023-06-01",
-    "num_hidden_layers": null,
+    "num_hidden_layers": 6,
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 32017,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -59642,7 +59642,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -59650,13 +59650,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 16384,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -59671,7 +59671,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -59679,13 +59679,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 16384,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -59787,7 +59787,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1026,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -59795,13 +59795,13 @@
     "hf_downloads": 87,
     "hf_likes": 0,
     "release_date": "2023-02-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 51416,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -59858,7 +59858,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 32768,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -60107,7 +60107,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -60115,12 +60115,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 14,
+    "num_key_value_heads": 14,
+    "head_dim": 64,
+    "hidden_size": 896,
+    "vocab_size": 16001,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -60442,7 +60442,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -60450,13 +60450,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 25,
+    "num_attention_heads": 10,
+    "num_key_value_heads": 5,
+    "head_dim": 64,
+    "hidden_size": 640,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 1728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -60749,7 +60749,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -60757,13 +60757,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 9,
+    "num_key_value_heads": 3,
+    "head_dim": 64,
+    "hidden_size": 576,
+    "vocab_size": 49152,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -62247,7 +62247,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 50257,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -63156,7 +63156,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -63164,13 +63164,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 9,
+    "num_key_value_heads": 3,
+    "head_dim": 64,
+    "hidden_size": 576,
+    "vocab_size": 49153,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -63327,7 +63327,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 3072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -63335,13 +63335,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 80,
+    "hidden_size": 640,
+    "vocab_size": 32768,
+    "moe_intermediate_size": 1920,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "bananamind-community-license-1.0"
@@ -63927,8 +63927,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -63962,8 +63962,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -63994,8 +63994,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64029,8 +64029,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64061,8 +64061,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64093,8 +64093,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64125,8 +64125,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64157,8 +64157,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -64188,8 +64188,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -64219,8 +64219,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64254,8 +64254,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64289,8 +64289,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64324,8 +64324,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64356,8 +64356,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64388,8 +64388,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64424,8 +64424,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64577,8 +64577,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -64611,8 +64611,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64646,8 +64646,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64678,8 +64678,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64710,8 +64710,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64742,8 +64742,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64777,8 +64777,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64809,8 +64809,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64844,8 +64844,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64876,8 +64876,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64908,8 +64908,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64940,8 +64940,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -64972,8 +64972,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -65004,8 +65004,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -65039,8 +65039,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 768,
+    "vocab_size": 81,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -65293,7 +65293,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -65301,13 +65301,13 @@
     "hf_downloads": 71,
     "hf_likes": 5,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 82000,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -65764,7 +65764,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8194,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -65775,13 +65775,13 @@
     "hf_downloads": 67,
     "hf_likes": 8,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 52009,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -65795,7 +65795,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -65803,13 +65803,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 52,
+    "hidden_size": 832,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 2496,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -66301,7 +66301,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -66309,13 +66309,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -66543,7 +66543,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -66554,13 +66554,13 @@
     "hf_downloads": 87,
     "hf_likes": 8,
     "release_date": "2023-07-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 23236,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "afl-3.0"
@@ -66883,13 +66883,13 @@
     "hf_downloads": 78,
     "hf_likes": 0,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
     "vocab_size": null,
-    "moe_intermediate_size": null,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -68457,7 +68457,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -68468,13 +68468,13 @@
     "hf_downloads": 88,
     "hf_likes": 0,
     "release_date": "2023-05-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 119547,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "afl-3.0"
@@ -68666,7 +68666,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -68674,13 +68674,13 @@
     "hf_downloads": 2409,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -68695,7 +68695,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -68703,13 +68703,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -68724,7 +68724,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -68732,13 +68732,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -68871,7 +68871,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -68879,13 +68879,13 @@
     "hf_downloads": 4018,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -68913,8 +68913,8 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "vocab_size": 12,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -69221,7 +69221,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -69229,13 +69229,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -69364,7 +69364,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -69375,13 +69375,13 @@
     "hf_downloads": 100,
     "hf_likes": 34,
     "release_date": "2023-12-25",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 35840,
+    "moe_intermediate_size": 4096,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -70158,7 +70158,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -70166,13 +70166,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 48,
+    "hidden_size": 384,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -74311,7 +74311,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -74319,13 +74319,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -74340,7 +74340,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -74348,13 +74348,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -74369,7 +74369,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -74377,13 +74377,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -74456,7 +74456,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -74464,13 +74464,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 35,
+    "num_attention_heads": 6,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 768,
+    "vocab_size": 16384,
+    "moe_intermediate_size": 1920,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -74760,7 +74760,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 32128,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -74957,7 +74957,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -74965,13 +74965,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -75298,7 +75298,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -75306,13 +75306,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 14,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 2560,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -75327,7 +75327,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -75335,13 +75335,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 14,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 2560,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -75689,7 +75689,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -75697,13 +75697,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 130560,
+    "moe_intermediate_size": 4608,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -76491,7 +76491,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -76499,13 +76499,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -76520,7 +76520,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -76532,13 +76532,13 @@
     "hf_downloads": 157,
     "hf_likes": 9,
     "release_date": "2026-01-06",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -76553,7 +76553,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -76561,13 +76561,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -76582,7 +76582,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -76590,13 +76590,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -77327,7 +77327,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -77337,13 +77337,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 5,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -77755,7 +77755,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -77763,13 +77763,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 18,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 640,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -77784,7 +77784,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -77795,13 +77795,13 @@
     "hf_downloads": 236,
     "hf_likes": 2,
     "release_date": "2026-08-06",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 18,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 640,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -77844,7 +77844,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -77854,13 +77854,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 18,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 640,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -77874,7 +77874,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -77882,13 +77882,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 18,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 640,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -79098,7 +79098,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -79118,13 +79118,13 @@
     "hf_downloads": 52,
     "hf_likes": 15,
     "release_date": "2023-08-11",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 250002,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -79268,7 +79268,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -79281,13 +79281,13 @@
     "hf_downloads": 81,
     "hf_likes": 0,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 250002,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -79302,7 +79302,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -79315,13 +79315,13 @@
     "hf_downloads": 82,
     "hf_likes": 0,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 250002,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -79336,7 +79336,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -79349,13 +79349,13 @@
     "hf_downloads": 25,
     "hf_likes": 0,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 250002,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -79427,7 +79427,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 514,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -79438,13 +79438,13 @@
     "hf_downloads": 28,
     "hf_likes": 0,
     "release_date": "2023-06-19",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 250002,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -80194,7 +80194,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -80205,13 +80205,13 @@
     "hf_downloads": 87,
     "hf_likes": 7,
     "release_date": "2023-09-21",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 12,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 21128,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "afl-3.0"
@@ -80709,7 +80709,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 16384,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -81047,13 +81047,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 64,
+    "hidden_size": 1024,
     "vocab_size": null,
-    "moe_intermediate_size": null,
+    "moe_intermediate_size": 4096,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -81754,13 +81754,13 @@
     "hf_downloads": 87,
     "hf_likes": 0,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 32,
+    "moe_intermediate_size": 4096,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -81872,13 +81872,13 @@
     "hf_downloads": 15,
     "hf_likes": 1,
     "release_date": "2022-03-02",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 32,
+    "moe_intermediate_size": 4096,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -82295,7 +82295,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -82303,13 +82303,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -82324,7 +82324,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -82332,13 +82332,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -82353,7 +82353,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -82361,13 +82361,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -82382,7 +82382,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -82390,13 +82390,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -82524,7 +82524,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -82535,13 +82535,13 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-05-19",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 28895,
+    "moe_intermediate_size": 4096,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -83064,7 +83064,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -83075,13 +83075,13 @@
     "hf_downloads": 91,
     "hf_likes": 2,
     "release_date": "2023-04-26",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 4096,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -83096,7 +83096,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -83104,13 +83104,13 @@
     "hf_downloads": 90,
     "hf_likes": 0,
     "release_date": "2023-10-13",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 30522,
+    "moe_intermediate_size": 4096,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -83378,7 +83378,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -83386,13 +83386,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 768,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -83937,7 +83937,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -83945,13 +83945,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -84589,7 +84589,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -84597,13 +84597,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 6656,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -84647,7 +84647,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -84655,13 +84655,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 6656,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -84705,7 +84705,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -84713,13 +84713,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 6656,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0",
@@ -84740,7 +84740,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -84748,13 +84748,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 6656,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -85422,7 +85422,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 50258,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -85718,7 +85718,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -85726,13 +85726,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 18,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 103424,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -85747,7 +85747,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -85755,13 +85755,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 18,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 103424,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -86253,7 +86253,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -86261,13 +86261,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 26,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 1152,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 6912,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -86504,7 +86504,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -86512,13 +86512,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 15,
+    "num_key_value_heads": 5,
+    "head_dim": 64,
+    "hidden_size": 960,
+    "vocab_size": 49153,
+    "moe_intermediate_size": 2560,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -87806,7 +87806,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 250100,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -88398,7 +88398,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -88409,13 +88409,13 @@
     "hf_downloads": 91,
     "hf_likes": 20,
     "release_date": "2025-03-23",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 50370,
+    "moe_intermediate_size": 2624,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -89022,7 +89022,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -89030,13 +89030,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm-open-license-v1.0"
@@ -89051,7 +89051,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -89059,13 +89059,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -89201,7 +89201,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 604,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -89520,13 +89520,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 15,
+    "num_key_value_heads": 5,
+    "head_dim": 64,
+    "hidden_size": 960,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 2560,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -89555,13 +89555,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 15,
+    "num_key_value_heads": 5,
+    "head_dim": 64,
+    "hidden_size": 960,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 2560,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -89576,7 +89576,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -89584,13 +89584,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 15,
+    "num_key_value_heads": 5,
+    "head_dim": 64,
+    "hidden_size": 960,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 2560,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -89700,13 +89700,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 20,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -89993,7 +89993,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -90003,13 +90003,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 6656,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -90892,7 +90892,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -90900,13 +90900,13 @@
     "hf_downloads": 10906,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 4,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 1024,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -91562,7 +91562,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -91570,13 +91570,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 512,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -92195,7 +92195,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -92203,13 +92203,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 14,
+    "num_key_value_heads": 2,
+    "head_dim": 64,
+    "hidden_size": 896,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 4864,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -92340,7 +92340,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -92348,13 +92348,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 14,
+    "num_key_value_heads": 2,
+    "head_dim": 64,
+    "hidden_size": 896,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 4864,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -92369,7 +92369,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [
       "tool_use"
@@ -92379,13 +92379,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 14,
+    "num_key_value_heads": 2,
+    "head_dim": 64,
+    "hidden_size": 896,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 4864,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -92400,7 +92400,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -92410,13 +92410,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 14,
+    "num_key_value_heads": 2,
+    "head_dim": 64,
+    "hidden_size": 896,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 4864,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -95552,7 +95552,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -95560,13 +95560,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 155776,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -98089,7 +98089,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -98097,13 +98097,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 14,
+    "num_key_value_heads": 2,
+    "head_dim": 64,
+    "hidden_size": 896,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 4864,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -98155,7 +98155,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Text embeddings for RAG",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -98163,13 +98163,13 @@
     "hf_downloads": 29,
     "hf_likes": 1,
     "release_date": "2026-08-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 151669,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -98755,7 +98755,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -98763,13 +98763,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "hamo-rail-s-1.0",
@@ -98988,7 +98988,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -98998,13 +98998,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -99056,7 +99056,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -99064,13 +99064,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -99279,7 +99279,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose",
     "capabilities": [],
     "languages": [
@@ -99290,13 +99290,13 @@
     "hf_downloads": 21,
     "hf_likes": 2,
     "release_date": "2026-08-11",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -100270,7 +100270,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -100278,13 +100278,13 @@
     "hf_downloads": 4163,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 14,
+    "num_key_value_heads": 2,
+    "head_dim": 64,
+    "hidden_size": 896,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 4864,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -101404,7 +101404,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -101412,13 +101412,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 2816,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -102251,7 +102251,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 12288,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -102262,13 +102262,13 @@
     "hf_downloads": 287,
     "hf_likes": 10,
     "release_date": "2026-08-03",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 90,
+    "hidden_size": 1440,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 3280,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -102283,7 +102283,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 12288,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -102291,13 +102291,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 90,
+    "hidden_size": 1440,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 3280,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -102831,7 +102831,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -102839,13 +102839,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 80,
+    "hidden_size": 1280,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 4480,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -104689,7 +104689,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -104697,13 +104697,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 1024,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 3584,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -105137,7 +105137,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -105145,13 +105145,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -105781,7 +105781,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 50257,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -106107,7 +106107,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 32128,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -106456,7 +106456,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -106464,13 +106464,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 4,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 32064,
+    "moe_intermediate_size": 3200,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -106571,7 +106571,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 49152,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -107751,7 +107751,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -107759,13 +107759,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 1024,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 3584,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -108843,7 +108843,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -108853,13 +108853,13 @@
     "hf_downloads": 121,
     "hf_likes": 1,
     "release_date": "2026-08-03",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -108905,7 +108905,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -108913,13 +108913,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 166144,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -109084,7 +109084,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -109092,13 +109092,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 166144,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -109439,7 +109439,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -109447,13 +109447,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 50304,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -111438,7 +111438,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -111446,13 +111446,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 26,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 1152,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 6912,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -111467,7 +111467,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -111475,13 +111475,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 26,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 1152,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 6912,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -111496,7 +111496,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -111504,13 +111504,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 26,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 1152,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 6912,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -111525,7 +111525,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -111533,13 +111533,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 26,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 1152,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 6912,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -112127,7 +112127,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8388608,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -112135,13 +112135,13 @@
     "hf_downloads": 1123,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 26,
+    "num_attention_heads": 4,
+    "num_key_value_heads": 1,
+    "head_dim": 288,
+    "hidden_size": 1152,
+    "vocab_size": 262149,
+    "moe_intermediate_size": 6912,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "osl-3.0"
@@ -112911,7 +112911,7 @@
     "min_vram_gb": 0.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -112922,14 +112922,14 @@
     "hf_downloads": 2340,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -113713,7 +113713,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 50432,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -114280,7 +114280,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -114288,13 +114288,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 130560,
+    "moe_intermediate_size": 4608,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -114309,7 +114309,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [
       "tool_use"
@@ -114322,13 +114322,13 @@
     "hf_downloads": 178,
     "hf_likes": 5,
     "release_date": "2026-05-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 130560,
+    "moe_intermediate_size": 4608,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -114343,7 +114343,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -114351,13 +114351,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 130560,
+    "moe_intermediate_size": 4608,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -114458,7 +114458,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -114466,13 +114466,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 130560,
+    "moe_intermediate_size": 4608,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -115383,7 +115383,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -115391,13 +115391,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 5632,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -115440,7 +115440,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -115448,13 +115448,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 5632,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -115468,7 +115468,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -115476,13 +115476,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 5632,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -116211,13 +116211,13 @@
     "hf_downloads": 445,
     "hf_likes": 64,
     "release_date": "2024-01-21",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 32002,
+    "moe_intermediate_size": 5632,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -117121,7 +117121,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 256000,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true
@@ -117425,7 +117425,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -117435,13 +117435,13 @@
     "hf_downloads": 1370,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -118428,7 +118428,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -118436,13 +118436,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -118457,7 +118457,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -118469,13 +118469,13 @@
     "hf_downloads": 186,
     "hf_likes": 9,
     "release_date": "2026-01-06",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -118490,7 +118490,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -118498,13 +118498,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -118519,7 +118519,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -118527,13 +118527,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -118927,7 +118927,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -118935,13 +118935,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 166144,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -120471,7 +120471,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -120479,13 +120479,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -120531,7 +120531,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -120539,13 +120539,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.2"
@@ -120647,7 +120647,7 @@
     "min_vram_gb": 0.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -120655,13 +120655,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -122237,13 +122237,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 14,
+    "num_key_value_heads": 14,
+    "head_dim": 73,
+    "hidden_size": 1024,
+    "vocab_size": 163840,
+    "moe_intermediate_size": 448,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "kimi-k3-license"
@@ -122298,7 +122298,7 @@
     "min_vram_gb": 0.7,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "Text embeddings for RAG",
     "capabilities": [
       "tool_use"
@@ -122308,13 +122308,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151665,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -122574,7 +122574,7 @@
     "min_vram_gb": 0.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -122582,13 +122582,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 1280,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 5760,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "kanana-open-license"
@@ -123429,7 +123429,7 @@
     "min_vram_gb": 0.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -123437,13 +123437,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -124242,12 +124242,12 @@
     "hf_downloads": 6635,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
+    "num_hidden_layers": 24,
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 2048,
+    "vocab_size": 32000,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -124878,7 +124878,7 @@
     "min_vram_gb": 0.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 134217728,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -124886,13 +124886,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 5,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 16,
+    "head_dim": 64,
+    "hidden_size": 4096,
+    "vocab_size": 201024,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -126489,7 +126489,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -126497,13 +126497,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -126526,13 +126526,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -127199,7 +127199,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -127207,13 +127207,13 @@
     "hf_downloads": 1083,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -127234,7 +127234,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -127242,13 +127242,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.2",
@@ -127269,7 +127269,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -127277,13 +127277,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.2",
@@ -127304,7 +127304,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -127312,13 +127312,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "gguf_sources": [
@@ -128174,7 +128174,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -128182,13 +128182,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 25,
+    "num_key_value_heads": 5,
+    "head_dim": 64,
+    "hidden_size": 1600,
+    "vocab_size": 32001,
+    "moe_intermediate_size": 5504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "nvidia-open-model-license"
@@ -128240,13 +128240,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
+    "num_hidden_layers": 24,
     "num_attention_heads": null,
     "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -129061,7 +129061,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -129071,13 +129071,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 8960,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "gguf_sources": [
@@ -129159,7 +129159,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -129167,13 +129167,13 @@
     "hf_downloads": 7930,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 8960,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "seallms",
@@ -129268,7 +129268,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -129278,13 +129278,13 @@
     "hf_downloads": 1929,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 8960,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -129344,7 +129344,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -129354,13 +129354,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 8960,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -129375,7 +129375,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -129385,13 +129385,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 8960,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -129406,7 +129406,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -129416,13 +129416,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 8960,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -129437,7 +129437,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -129445,13 +129445,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 8960,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "katanemo-research",
@@ -131532,7 +131532,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -131548,13 +131548,13 @@
     "hf_downloads": 389,
     "hf_likes": 11,
     "release_date": "2025-08-26",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -132248,13 +132248,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 32064,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -132268,7 +132268,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -132278,13 +132278,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -132732,7 +132732,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -132740,13 +132740,13 @@
     "hf_downloads": 1299,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 59264,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "glm-4",
@@ -133464,7 +133464,7 @@
     "min_vram_gb": 0.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -133472,13 +133472,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 196608,
+    "moe_intermediate_size": 1280,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "upstage-solar-license"
@@ -134347,7 +134347,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -134355,13 +134355,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 49152,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "webai-non-commercial-license-ver.-1.0"
@@ -134439,7 +134439,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -134447,13 +134447,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 49153,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -134731,7 +134731,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -134741,13 +134741,13 @@
     "hf_downloads": 45,
     "hf_likes": 2,
     "release_date": "2026-04-13",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -137012,7 +137012,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -137020,13 +137020,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 8960,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -137041,7 +137041,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -137051,13 +137051,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 8960,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -137099,7 +137099,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -137107,13 +137107,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 8960,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -137714,7 +137714,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -137722,13 +137722,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 120818,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -137742,7 +137742,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -137750,13 +137750,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2880,
+    "vocab_size": 201088,
+    "moe_intermediate_size": 2880,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -137771,7 +137771,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -137779,13 +137779,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2880,
+    "vocab_size": 201088,
+    "moe_intermediate_size": 2880,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -137829,7 +137829,7 @@
     "min_vram_gb": 1.9,
     "quantization": "GPTQ-Int8",
     "format": "gptq",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -137837,13 +137837,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 166144,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -137916,7 +137916,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -137926,13 +137926,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 5,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -138004,7 +138004,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -138012,13 +138012,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 96,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 8320,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -138380,7 +138380,7 @@
     "min_vram_gb": 0.9,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -138388,13 +138388,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 1792,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -139408,7 +139408,7 @@
     "min_vram_gb": 1.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -139418,13 +139418,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -140188,7 +140188,7 @@
     "min_vram_gb": 1.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -140196,13 +140196,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -141052,7 +141052,7 @@
     "min_vram_gb": 1.1,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -141062,13 +141062,13 @@
     "hf_downloads": 2118,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -141083,7 +141083,7 @@
     "min_vram_gb": 1.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -141091,13 +141091,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 1792,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8064,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -141170,7 +141170,7 @@
     "min_vram_gb": 1.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -141178,13 +141178,13 @@
     "hf_downloads": 2829,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 1792,
+    "vocab_size": 128259,
+    "moe_intermediate_size": 8064,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -141875,7 +141875,7 @@
     "min_vram_gb": 1.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -141885,13 +141885,13 @@
     "hf_downloads": 29211,
     "hf_likes": 112,
     "release_date": "2026-03-14",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -142308,7 +142308,7 @@
     "min_vram_gb": 1.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -142318,13 +142318,13 @@
     "hf_downloads": 165,
     "hf_likes": 1,
     "release_date": "2026-08-07",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -142854,7 +142854,7 @@
     "min_vram_gb": 1.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -142862,13 +142862,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -142883,7 +142883,7 @@
     "min_vram_gb": 1.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -142893,13 +142893,13 @@
     "hf_downloads": 81,
     "hf_likes": 2,
     "release_date": "2026-08-03",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -144710,7 +144710,7 @@
     "min_vram_gb": 1.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -144720,13 +144720,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -144741,7 +144741,7 @@
     "min_vram_gb": 1.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -144751,13 +144751,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -144772,7 +144772,7 @@
     "min_vram_gb": 1.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -144782,13 +144782,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 13824,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -144803,7 +144803,7 @@
     "min_vram_gb": 1.2,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -144813,13 +144813,13 @@
     "hf_downloads": 3185,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 13824,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -145430,7 +145430,7 @@
     "min_vram_gb": 1.2,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -145438,13 +145438,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 1792,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -146353,7 +146353,7 @@
     "min_vram_gb": 1.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -146361,13 +146361,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 120832,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -146568,7 +146568,7 @@
     "min_vram_gb": 1.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -146576,13 +146576,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 18,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 16384,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -147789,13 +147789,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 64,
+    "hidden_size": 1024,
+    "vocab_size": 49152,
+    "moe_intermediate_size": 512,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -148280,7 +148280,7 @@
     "min_vram_gb": 1.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -148291,13 +148291,13 @@
     "hf_downloads": 11751,
     "hf_likes": 53,
     "release_date": "2024-07-31",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 26,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2304,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -148312,7 +148312,7 @@
     "min_vram_gb": 1.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -148324,13 +148324,13 @@
     "hf_downloads": 363,
     "hf_likes": 17,
     "release_date": "2024-08-18",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 26,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2304,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -149316,7 +149316,7 @@
     "min_vram_gb": 1.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -149324,13 +149324,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -149345,7 +149345,7 @@
     "min_vram_gb": 1.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -149353,13 +149353,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -149374,7 +149374,7 @@
     "min_vram_gb": 1.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -149382,13 +149382,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -149553,7 +149553,7 @@
     "min_vram_gb": 1.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -149561,13 +149561,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -149611,7 +149611,7 @@
     "min_vram_gb": 1.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -149619,13 +149619,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -150710,7 +150710,7 @@
     "min_vram_gb": 1.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -150718,13 +150718,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 80,
+    "hidden_size": 2560,
+    "vocab_size": 51200,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "stem.ai.mtl"
@@ -150836,7 +150836,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 51200,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -150888,13 +150888,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 80,
+    "hidden_size": 2560,
+    "vocab_size": 50280,
+    "moe_intermediate_size": 6912,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -151134,7 +151134,7 @@
     "min_vram_gb": 1.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16384,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -151142,13 +151142,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 80,
+    "hidden_size": 2560,
+    "vocab_size": 50304,
+    "moe_intermediate_size": 6912,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -151282,7 +151282,7 @@
     "min_vram_gb": 1.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -151295,13 +151295,13 @@
     "hf_downloads": 98960,
     "hf_likes": 2,
     "release_date": "2025-08-27",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -151406,7 +151406,7 @@
     "min_vram_gb": 1.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -151414,13 +151414,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 7168,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -151802,7 +151802,7 @@
     "min_vram_gb": 1.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "vision",
@@ -151813,13 +151813,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -152105,7 +152105,7 @@
     "min_vram_gb": 1.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -152113,13 +152113,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 88,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 2688,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "nvidia-nemotron-open-model-license"
@@ -152535,7 +152535,7 @@
     "min_vram_gb": 1.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -152543,13 +152543,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 2816,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 704,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -152977,7 +152977,7 @@
     "min_vram_gb": 1.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 524288,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -152985,13 +152985,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 1,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 112,
+    "hidden_size": 7168,
+    "vocab_size": 163840,
+    "moe_intermediate_size": 18432,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -153319,7 +153319,7 @@
     "min_vram_gb": 1.6,
     "quantization": "AutoRound-4bit",
     "format": "autoround",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -153330,13 +153330,13 @@
     "hf_downloads": 20,
     "hf_likes": 2,
     "release_date": "2026-08-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -153351,7 +153351,7 @@
     "min_vram_gb": 1.6,
     "quantization": "AutoRound-4bit",
     "format": "autoround",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -153361,13 +153361,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -154313,7 +154313,7 @@
     "min_vram_gb": 1.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 65536,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -154321,13 +154321,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -155046,7 +155046,7 @@
     "min_vram_gb": 1.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -155054,13 +155054,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -155166,7 +155166,7 @@
     "min_vram_gb": 1.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -155174,13 +155174,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -155195,7 +155195,7 @@
     "min_vram_gb": 1.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -155203,13 +155203,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -155284,7 +155284,7 @@
     "min_vram_gb": 1.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -155292,13 +155292,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -158009,7 +158009,7 @@
     "min_vram_gb": 1.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1310720,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -158017,13 +158017,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "kanana-open-license"
@@ -158575,7 +158575,7 @@
     "min_vram_gb": 1.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -158585,13 +158585,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -158606,7 +158606,7 @@
     "min_vram_gb": 1.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -158614,13 +158614,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.2"
@@ -158635,7 +158635,7 @@
     "min_vram_gb": 1.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -158645,13 +158645,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -159704,7 +159704,7 @@
     "min_vram_gb": 1.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -159712,13 +159712,13 @@
     "hf_downloads": 17625,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 3072,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "falcon-llm-license",
@@ -159743,7 +159743,7 @@
     "min_vram_gb": 1.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -159751,13 +159751,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 3072,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -159983,7 +159983,7 @@
     "min_vram_gb": 1.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -159991,13 +159991,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 5,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -160012,7 +160012,7 @@
     "min_vram_gb": 1.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -160020,13 +160020,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 6,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 200064,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "nvidia-open-model-license"
@@ -160103,7 +160103,7 @@
     "min_vram_gb": 1.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -160111,13 +160111,13 @@
     "hf_downloads": 3247,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 1536,
+    "vocab_size": 49152,
+    "moe_intermediate_size": 512,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -160742,7 +160742,7 @@
     "min_vram_gb": 1.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -160750,13 +160750,13 @@
     "hf_downloads": 1823,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 26,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 100,
+    "hidden_size": 3200,
+    "vocab_size": 32002,
+    "moe_intermediate_size": 8640,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -160853,7 +160853,7 @@
     "min_vram_gb": 1.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -160863,13 +160863,13 @@
     "hf_downloads": 314,
     "hf_likes": 1,
     "release_date": "2026-07-31",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 12,
+    "num_attention_heads": 10,
+    "num_key_value_heads": 10,
+    "head_dim": 128,
+    "hidden_size": 1280,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 896,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -161598,7 +161598,7 @@
     "min_vram_gb": 1.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -161607,11 +161607,11 @@
     "hf_likes": 1,
     "release_date": "2023-04-16",
     "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 130528,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -162791,7 +162791,7 @@
     "min_vram_gb": 1.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -162799,13 +162799,13 @@
     "hf_downloads": 53087,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2560,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -163822,7 +163822,7 @@
     "min_vram_gb": 1.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -163830,13 +163830,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 26,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 100,
+    "hidden_size": 3200,
+    "vocab_size": 32003,
+    "moe_intermediate_size": 8640,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -164734,7 +164734,7 @@
     "min_vram_gb": 1.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -164742,14 +164742,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -165507,13 +165507,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 32768,
+    "moe_intermediate_size": 4096,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-4.0",
@@ -165826,7 +165826,7 @@
     "min_vram_gb": 1.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 202752,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -165834,13 +165834,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 47,
+    "num_attention_heads": 20,
+    "num_key_value_heads": 20,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -167284,7 +167284,7 @@
     "min_vram_gb": 2.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -167292,13 +167292,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 96,
+    "hidden_size": 3072,
+    "vocab_size": 32064,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -167321,13 +167321,13 @@
     "hf_downloads": 1038,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 96,
+    "hidden_size": 3072,
+    "vocab_size": 32064,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -167352,7 +167352,7 @@
     "min_vram_gb": 2.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -167360,13 +167360,13 @@
     "hf_downloads": 1755,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 96,
+    "hidden_size": 3072,
+    "vocab_size": 32064,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-sa-4.0",
@@ -168240,13 +168240,13 @@
     "hf_downloads": 578,
     "hf_likes": 3,
     "release_date": "2026-04-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 16384,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -168582,7 +168582,7 @@
     "min_vram_gb": 2.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -168590,13 +168590,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 200064,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "inherits-base-model-and-dataset-terms"
@@ -168611,7 +168611,7 @@
     "min_vram_gb": 2.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -168619,13 +168619,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 200064,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "inherits-base-model-and-dataset-terms",
@@ -168771,7 +168771,7 @@
     "min_vram_gb": 2.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "languages": [
@@ -168782,13 +168782,13 @@
     "hf_downloads": 89,
     "hf_likes": 4,
     "release_date": "2025-11-09",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -168976,7 +168976,7 @@
     "min_vram_gb": 2.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -168984,13 +168984,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 20,
+    "head_dim": 64,
+    "hidden_size": 2560,
+    "vocab_size": 200064,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -169469,7 +169469,7 @@
     "min_vram_gb": 2.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 65536,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -169477,13 +169477,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 20,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 166144,
+    "moe_intermediate_size": 10496,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -170161,7 +170161,7 @@
     "min_vram_gb": 2.0,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -170176,13 +170176,13 @@
     "hf_downloads": 315,
     "hf_likes": 3,
     "release_date": "2026-08-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -172060,7 +172060,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -172068,13 +172068,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -172148,7 +172148,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -172156,13 +172156,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -172177,7 +172177,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -172187,13 +172187,13 @@
     "hf_downloads": 23428,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -172214,7 +172214,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -172224,13 +172224,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -172334,7 +172334,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -172342,13 +172342,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -172393,7 +172393,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -172403,13 +172403,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -172424,7 +172424,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -172435,13 +172435,13 @@
     "hf_downloads": 496,
     "hf_likes": 19,
     "release_date": "2025-06-20",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -172456,7 +172456,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -172466,13 +172466,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -172487,7 +172487,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -172497,13 +172497,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -172518,7 +172518,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -172526,13 +172526,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -172547,7 +172547,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -172555,13 +172555,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -172605,7 +172605,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -172613,13 +172613,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -172694,7 +172694,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [
       "tool_use"
@@ -172704,13 +172704,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -172994,7 +172994,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -173004,13 +173004,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -173176,7 +173176,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -173184,13 +173184,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -176904,7 +176904,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -176912,13 +176912,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 166144,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -176962,7 +176962,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -176970,13 +176970,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 22,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 166144,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -177135,7 +177135,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -177143,13 +177143,13 @@
     "hf_downloads": 9732,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 34,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 8640,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -177163,7 +177163,7 @@
     "min_vram_gb": 2.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -177171,13 +177171,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 34,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 8640,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "nvidia-open-model-license-agreement"
@@ -177250,7 +177250,7 @@
     "min_vram_gb": 2.1,
     "quantization": "AutoRound-4bit",
     "format": "autoround",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -177261,14 +177261,14 @@
     "hf_downloads": 983,
     "hf_likes": 4,
     "release_date": "2026-07-23",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "is_moe": true,
     "num_experts": 256,
@@ -177648,7 +177648,7 @@
     "min_vram_gb": 2.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -177656,13 +177656,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "agpl-3.0"
@@ -177677,7 +177677,7 @@
     "min_vram_gb": 2.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -177685,13 +177685,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -177885,7 +177885,7 @@
     "min_vram_gb": 2.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -177895,13 +177895,13 @@
     "hf_downloads": 361,
     "hf_likes": 1,
     "release_date": "2026-08-05",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -178127,7 +178127,7 @@
     "min_vram_gb": 2.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -178135,12 +178135,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1536,
+    "vocab_size": 151936,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -178307,7 +178307,7 @@
     "min_vram_gb": 2.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -178315,13 +178315,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 34,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 262145,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -178336,7 +178336,7 @@
     "min_vram_gb": 2.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -178344,13 +178344,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 34,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 262145,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -178567,7 +178567,7 @@
     "min_vram_gb": 2.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -178577,13 +178577,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 34,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 262208,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -178657,7 +178657,7 @@
     "min_vram_gb": 2.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -178665,13 +178665,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 34,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 262208,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -178686,7 +178686,7 @@
     "min_vram_gb": 2.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "languages": [
@@ -178697,13 +178697,13 @@
     "hf_downloads": 105,
     "hf_likes": 1,
     "release_date": "2025-04-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 34,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 262208,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -178775,7 +178775,7 @@
     "min_vram_gb": 2.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -178783,13 +178783,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 34,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 262208,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -180941,7 +180941,7 @@
     "min_vram_gb": 2.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -180949,13 +180949,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -181028,7 +181028,7 @@
     "min_vram_gb": 2.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -181038,13 +181038,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -181059,7 +181059,7 @@
     "min_vram_gb": 2.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -181069,13 +181069,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -182860,7 +182860,7 @@
     "min_vram_gb": 2.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 524288,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -182868,13 +182868,13 @@
     "hf_downloads": 3075,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "nvidia-open-model-license",
@@ -182952,7 +182952,7 @@
     "min_vram_gb": 2.3,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -182966,13 +182966,13 @@
     "hf_downloads": 666,
     "hf_likes": 8,
     "release_date": "2026-04-06",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 2816,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 704,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -183478,7 +183478,7 @@
     "min_vram_gb": 2.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -183489,13 +183489,13 @@
     "hf_downloads": 337,
     "hf_likes": 14,
     "release_date": "2026-03-07",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -183510,7 +183510,7 @@
     "min_vram_gb": 2.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -183525,13 +183525,13 @@
     "hf_downloads": 567,
     "hf_likes": 22,
     "release_date": "2026-03-10",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -183578,7 +183578,7 @@
     "min_vram_gb": 2.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -183586,13 +183586,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -185270,7 +185270,7 @@
     "min_vram_gb": 2.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1310720,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -185278,13 +185278,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 20,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 128005,
+    "moe_intermediate_size": 18432,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -186157,7 +186157,7 @@
     "min_vram_gb": 2.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -186168,13 +186168,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -187827,7 +187827,7 @@
     "min_vram_gb": 2.5,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -187838,13 +187838,13 @@
     "hf_downloads": 438,
     "hf_likes": 1,
     "release_date": "2026-08-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -188389,7 +188389,7 @@
     "min_vram_gb": 2.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -188399,13 +188399,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -189105,7 +189105,7 @@
     "min_vram_gb": 2.6,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -189113,13 +189113,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -189320,7 +189320,7 @@
     "min_vram_gb": 2.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -189328,13 +189328,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 52,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 2688,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 1856,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -189349,7 +189349,7 @@
     "min_vram_gb": 2.6,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -189360,13 +189360,13 @@
     "hf_downloads": 386,
     "hf_likes": 1,
     "release_date": "2026-08-14",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -189381,7 +189381,7 @@
     "min_vram_gb": 2.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -189392,13 +189392,13 @@
     "hf_downloads": 2195,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151674,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -189573,7 +189573,7 @@
     "min_vram_gb": 2.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -189586,13 +189586,13 @@
     "hf_downloads": 350,
     "hf_likes": 46,
     "release_date": "2026-07-07",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 35,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 1536,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -189748,7 +189748,7 @@
     "min_vram_gb": 2.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -189756,13 +189756,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 35,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 1536,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -189968,7 +189968,7 @@
     "min_vram_gb": 2.6,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -189976,13 +189976,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -189997,7 +189997,7 @@
     "min_vram_gb": 2.6,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -190005,13 +190005,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -190429,7 +190429,7 @@
     "min_vram_gb": 2.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -190437,13 +190437,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 35,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 1,
+    "head_dim": 256,
+    "hidden_size": 1536,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -190962,7 +190962,7 @@
     "min_vram_gb": 2.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -190972,13 +190972,13 @@
     "hf_downloads": 3272,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -191433,7 +191433,7 @@
     "min_vram_gb": 2.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -191448,13 +191448,13 @@
     "hf_downloads": 98,
     "hf_likes": 1,
     "release_date": "2026-07-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -191877,7 +191877,7 @@
     "min_vram_gb": 2.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -191885,14 +191885,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "openmdw-1.1"
   },
@@ -192156,7 +192156,7 @@
     "min_vram_gb": 2.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -192167,13 +192167,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151674,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -192340,7 +192340,7 @@
     "min_vram_gb": 2.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -192348,13 +192348,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -192413,7 +192413,7 @@
     "min_vram_gb": 2.7,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -192426,13 +192426,13 @@
     "hf_downloads": 954,
     "hf_likes": 1,
     "release_date": "2026-07-18",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -192536,7 +192536,7 @@
     "min_vram_gb": 2.8,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -192549,14 +192549,14 @@
     "hf_downloads": 2475,
     "hf_likes": 2,
     "release_date": "2026-07-29",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -193364,7 +193364,7 @@
     "min_vram_gb": 2.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -193372,13 +193372,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -193393,7 +193393,7 @@
     "min_vram_gb": 2.9,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -193401,13 +193401,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 10752,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -193711,7 +193711,7 @@
     "min_vram_gb": 3.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -193719,13 +193719,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 121,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 512,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -193829,7 +193829,7 @@
     "min_vram_gb": 3.0,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -193840,13 +193840,13 @@
     "hf_downloads": 532,
     "hf_likes": 2,
     "release_date": "2026-08-14",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -193993,7 +193993,7 @@
     "min_vram_gb": 3.0,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -194003,14 +194003,14 @@
     "hf_downloads": 5285,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -194204,7 +194204,7 @@
     "min_vram_gb": 3.0,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -194214,13 +194214,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -194458,7 +194458,7 @@
     "min_vram_gb": 3.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -194468,13 +194468,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 55,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -194495,7 +194495,7 @@
     "min_vram_gb": 3.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 419430,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -194505,13 +194505,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 55,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -194532,7 +194532,7 @@
     "min_vram_gb": 3.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -194542,13 +194542,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 55,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -195802,7 +195802,7 @@
     "min_vram_gb": 3.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -195810,13 +195810,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 56,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 200192,
+    "moe_intermediate_size": 256,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -196172,7 +196172,7 @@
     "min_vram_gb": 3.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose",
     "capabilities": [],
     "pipeline_tag": "feature-extraction",
@@ -196181,11 +196181,11 @@
     "hf_likes": 0,
     "release_date": "2023-04-16",
     "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 130528,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -196493,7 +196493,7 @@
     "min_vram_gb": 3.2,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -196501,13 +196501,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 52,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 6656,
+    "vocab_size": 202048,
+    "moe_intermediate_size": 19968,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -196732,7 +196732,7 @@
     "min_vram_gb": 3.2,
     "quantization": "AutoRound-4bit",
     "format": "autoround",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -196742,13 +196742,13 @@
     "hf_downloads": 12277,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -197833,7 +197833,7 @@
     "min_vram_gb": 3.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -197841,13 +197841,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32064,
+    "moe_intermediate_size": 6400,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -197927,13 +197927,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 49,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -198257,7 +198257,7 @@
     "min_vram_gb": 3.4,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -198265,13 +198265,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "wtfpl"
@@ -199008,7 +199008,7 @@
     "min_vram_gb": 3.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16384,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -199016,13 +199016,13 @@
     "hf_downloads": 3585,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama2"
@@ -200019,7 +200019,7 @@
     "min_vram_gb": 3.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -200027,13 +200027,13 @@
     "hf_downloads": 2506,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other",
@@ -201788,7 +201788,7 @@
     "min_vram_gb": 3.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 65536,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "languages": [
@@ -201799,13 +201799,13 @@
     "hf_downloads": 13265,
     "hf_likes": 135,
     "release_date": "2024-02-19",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32256,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -202134,7 +202134,7 @@
     "min_vram_gb": 3.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -202142,13 +202142,13 @@
     "hf_downloads": 1107,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 41510,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-sa-4.0",
@@ -202268,13 +202268,13 @@
     "hf_downloads": 1929,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 46305,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama2"
@@ -202981,13 +202981,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 102400,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -204035,7 +204035,7 @@
     "min_vram_gb": 3.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -204045,13 +204045,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 6144,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -204720,13 +204720,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -205126,7 +205126,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -205134,13 +205134,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 67,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -205189,7 +205189,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -205199,13 +205199,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -205664,8 +205664,8 @@
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "hidden_size": 4544,
+    "vocab_size": 65025,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -206272,7 +206272,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206280,13 +206280,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -206300,7 +206300,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206308,13 +206308,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -206329,7 +206329,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206337,13 +206337,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -206358,7 +206358,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206366,13 +206366,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -206386,7 +206386,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206394,13 +206394,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -206414,7 +206414,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206422,13 +206422,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -206471,7 +206471,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206479,13 +206479,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -206500,7 +206500,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206508,13 +206508,13 @@
     "hf_downloads": 98644,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -206529,7 +206529,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206537,13 +206537,13 @@
     "hf_downloads": 3818,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -206558,7 +206558,7 @@
     "min_vram_gb": 3.7,
     "quantization": "AWQ-4bit",
     "format": "awq",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206566,13 +206566,13 @@
     "hf_downloads": 119,
     "hf_likes": 5,
     "release_date": "2024-02-17",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -206644,7 +206644,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206652,13 +206652,13 @@
     "hf_downloads": 17955,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -206673,7 +206673,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206681,13 +206681,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -206702,7 +206702,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206710,13 +206710,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -206731,7 +206731,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -206743,13 +206743,13 @@
     "hf_downloads": 51,
     "hf_likes": 3,
     "release_date": "2024-03-26",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -206764,7 +206764,7 @@
     "min_vram_gb": 3.7,
     "quantization": "AWQ-4bit",
     "format": "awq",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206772,13 +206772,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -206792,7 +206792,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206800,13 +206800,13 @@
     "hf_downloads": 1204,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -206827,7 +206827,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206835,13 +206835,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -206856,7 +206856,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -206864,13 +206864,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -209576,7 +209576,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -209584,13 +209584,13 @@
     "hf_downloads": 2423,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32001,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -209777,7 +209777,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -209787,13 +209787,13 @@
     "hf_downloads": 1349,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32002,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -210598,7 +210598,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -210608,13 +210608,13 @@
     "hf_downloads": 4428,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32032,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -210900,7 +210900,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -210908,13 +210908,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32768,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -211458,7 +211458,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -211466,13 +211466,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32770,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -211595,7 +211595,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 65536,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -211603,13 +211603,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 92416,
+    "moe_intermediate_size": 13440,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "tongyi-qianwen-research"
@@ -212021,13 +212021,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
+    "num_hidden_layers": 64,
     "num_attention_heads": null,
     "num_key_value_heads": null,
     "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "hidden_size": 4096,
+    "vocab_size": 65024,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -212850,7 +212850,7 @@
     "min_vram_gb": 3.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -212858,13 +212858,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 100278,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -213032,7 +213032,7 @@
     "min_vram_gb": 3.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -213042,13 +213042,13 @@
     "hf_downloads": 509,
     "hf_likes": 2,
     "release_date": "2026-08-13",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 52,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 6656,
+    "vocab_size": 202048,
+    "moe_intermediate_size": 19968,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -213389,13 +213389,13 @@
     "hf_downloads": 1656,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 250680,
+    "moe_intermediate_size": 13440,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -213625,7 +213625,7 @@
     "min_vram_gb": 3.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -213633,13 +213633,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -213770,7 +213770,7 @@
     "min_vram_gb": 3.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -213778,13 +213778,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -213904,13 +213904,13 @@
     "hf_downloads": 2103,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -214072,7 +214072,7 @@
     "min_vram_gb": 3.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -214080,13 +214080,13 @@
     "hf_downloads": 1416,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 61952,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -214591,7 +214591,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -214599,13 +214599,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -214620,7 +214620,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -214628,13 +214628,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 67,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -214928,13 +214928,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -214949,7 +214949,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -214957,13 +214957,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 145152,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -215577,7 +215577,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -215585,13 +215585,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 44,
+    "num_attention_heads": 12,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 130048,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "falcon-llm-license"
@@ -215798,7 +215798,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -215808,13 +215808,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 151669,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -217692,7 +217692,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -217700,13 +217700,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -217899,7 +217899,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -217907,13 +217907,13 @@
     "hf_downloads": 1612,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -217938,7 +217938,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -217946,13 +217946,13 @@
     "hf_downloads": 33136,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "nvidia-open-model-license"
@@ -218027,7 +218027,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -218035,13 +218035,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -218148,7 +218148,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -218156,13 +218156,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -218176,7 +218176,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -218184,13 +218184,13 @@
     "hf_downloads": 1187,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -218215,7 +218215,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -218225,13 +218225,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -218245,7 +218245,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -218255,13 +218255,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -218276,7 +218276,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -218284,13 +218284,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -218305,7 +218305,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -218313,13 +218313,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "nvidia-open-model-license"
@@ -218334,7 +218334,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -218342,13 +218342,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -218392,7 +218392,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -218400,13 +218400,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-4.0"
@@ -218421,7 +218421,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -218429,13 +218429,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -218715,7 +218715,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -218723,13 +218723,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apple-amlr"
@@ -221815,7 +221815,7 @@
     "min_vram_gb": 3.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -221823,14 +221823,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -222495,7 +222495,7 @@
     "min_vram_gb": 4.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -222503,13 +222503,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -223556,7 +223556,7 @@
     "min_vram_gb": 4.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -223564,13 +223564,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151680,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -223585,7 +223585,7 @@
     "min_vram_gb": 4.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -223593,13 +223593,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151680,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -223614,7 +223614,7 @@
     "min_vram_gb": 4.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 65536,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -223622,13 +223622,13 @@
     "hf_downloads": 4045,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151680,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -224837,7 +224837,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -224845,13 +224845,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 8,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2560,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 10240,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -225522,7 +225522,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -225530,13 +225530,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 1024,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 384,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -225555,7 +225555,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -225565,13 +225565,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -225877,7 +225877,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -225894,13 +225894,13 @@
     "hf_downloads": 593,
     "hf_likes": 8,
     "release_date": "2026-08-01",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -228345,7 +228345,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -228353,13 +228353,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -228527,7 +228527,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [
       "tool_use"
@@ -228537,13 +228537,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -228558,7 +228558,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -228566,13 +228566,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -228624,7 +228624,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -228632,13 +228632,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -228681,7 +228681,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -228689,13 +228689,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -228741,7 +228741,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -228751,13 +228751,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -228772,7 +228772,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -228780,13 +228780,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -228896,7 +228896,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -228904,13 +228904,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -228924,7 +228924,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -228932,13 +228932,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -228952,7 +228952,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -228960,13 +228960,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -228981,7 +228981,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -228989,13 +228989,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -229010,7 +229010,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229018,13 +229018,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -229039,7 +229039,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229047,13 +229047,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -229068,7 +229068,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229076,13 +229076,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -229097,7 +229097,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229105,13 +229105,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -229126,7 +229126,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229134,13 +229134,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -229212,7 +229212,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -229222,13 +229222,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -229243,7 +229243,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229251,13 +229251,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -229272,7 +229272,7 @@
     "min_vram_gb": 4.1,
     "quantization": "AWQ-4bit",
     "format": "awq",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229280,13 +229280,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -229301,7 +229301,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -229311,13 +229311,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -229331,7 +229331,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229339,13 +229339,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.1"
@@ -229360,7 +229360,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229368,13 +229368,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -229389,7 +229389,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229397,13 +229397,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -229446,7 +229446,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229454,13 +229454,13 @@
     "hf_downloads": 239,
     "hf_likes": 6,
     "release_date": "2024-09-09",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -229506,7 +229506,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229514,13 +229514,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.1"
@@ -229535,7 +229535,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229543,13 +229543,13 @@
     "hf_downloads": 1229,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -229574,7 +229574,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229582,13 +229582,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -229602,7 +229602,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229610,13 +229610,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -229784,7 +229784,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -229792,13 +229792,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -236778,7 +236778,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -236786,13 +236786,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128258,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -236807,7 +236807,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -236815,13 +236815,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128258,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -238147,7 +238147,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2197815296,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -238157,13 +238157,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 129024,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -238453,7 +238453,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 524288,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -238461,13 +238461,13 @@
     "hf_downloads": 1061,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -238782,7 +238782,7 @@
     "min_vram_gb": 4.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -238791,12 +238791,12 @@
     "hf_likes": 0,
     "release_date": null,
     "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 7688,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -242186,7 +242186,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -242194,13 +242194,13 @@
     "hf_downloads": 5782,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -242221,7 +242221,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -242231,13 +242231,13 @@
     "hf_downloads": 1712,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "gguf_sources": [
@@ -242257,7 +242257,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 524288,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -242265,13 +242265,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -242315,7 +242315,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -242323,13 +242323,13 @@
     "hf_downloads": 1105,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "nvidia-open-model-license",
@@ -242350,7 +242350,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -242358,13 +242358,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -242416,13 +242416,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -242437,7 +242437,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -242447,13 +242447,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -242526,7 +242526,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -242534,13 +242534,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -242555,7 +242555,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -242563,13 +242563,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -242619,7 +242619,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -242629,13 +242629,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -242650,7 +242650,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "languages": [
@@ -242662,13 +242662,13 @@
     "hf_downloads": 40,
     "hf_likes": 14,
     "release_date": "2025-11-18",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -242958,7 +242958,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 524288,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -242966,13 +242966,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -245225,7 +245225,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -245233,13 +245233,13 @@
     "hf_downloads": 4993,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 149248,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.1",
@@ -246373,7 +246373,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -246381,13 +246381,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -247068,7 +247068,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -247076,13 +247076,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 155776,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -249035,7 +249035,7 @@
     "min_vram_gb": 4.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -249046,13 +249046,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 3584,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 18944,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -250016,7 +250016,7 @@
     "min_vram_gb": 4.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "vision"
@@ -250026,14 +250026,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -250541,7 +250541,7 @@
     "min_vram_gb": 4.3,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -250549,14 +250549,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "is_moe": true,
     "num_experts": 256,
@@ -251023,7 +251023,7 @@
     "min_vram_gb": 4.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -251031,13 +251031,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 1792,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -251284,7 +251284,7 @@
     "min_vram_gb": 4.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -251292,13 +251292,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 1792,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -251342,7 +251342,7 @@
     "min_vram_gb": 4.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -251350,13 +251350,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 1792,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -251371,7 +251371,7 @@
     "min_vram_gb": 4.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -251379,13 +251379,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 1792,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -251954,7 +251954,7 @@
     "min_vram_gb": 4.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -251962,13 +251962,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -252682,13 +252682,13 @@
     "hf_downloads": 1101,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 34,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 160,
+    "hidden_size": 5120,
+    "vocab_size": 50176,
+    "moe_intermediate_size": 12800,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -253205,7 +253205,7 @@
     "min_vram_gb": 4.4,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -253215,13 +253215,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -253329,7 +253329,7 @@
     "min_vram_gb": 4.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 65536,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "languages": [
@@ -253341,13 +253341,13 @@
     "hf_downloads": 162,
     "hf_likes": 12,
     "release_date": "2026-04-14",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 196608,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -253989,7 +253989,7 @@
     "min_vram_gb": 4.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -254003,13 +254003,13 @@
     "hf_downloads": 988,
     "hf_likes": 6,
     "release_date": "2026-01-29",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -254068,7 +254068,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 32128,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -254581,7 +254581,7 @@
     "min_vram_gb": 4.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -254591,13 +254591,13 @@
     "hf_downloads": 4594,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -254829,7 +254829,7 @@
     "min_vram_gb": 4.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -254839,13 +254839,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -254860,7 +254860,7 @@
     "min_vram_gb": 4.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "vision"
@@ -254870,13 +254870,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -256167,7 +256167,7 @@
     "min_vram_gb": 4.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -256175,13 +256175,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 12800,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -256456,7 +256456,7 @@
     "min_vram_gb": 4.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16384,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -256464,13 +256464,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 64000,
+    "moe_intermediate_size": 11008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -258929,7 +258929,7 @@
     "min_vram_gb": 4.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -258937,13 +258937,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "polyform-noncommercial-1.0.0"
@@ -258958,7 +258958,7 @@
     "min_vram_gb": 4.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -258966,13 +258966,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -259016,7 +259016,7 @@
     "min_vram_gb": 4.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -259024,13 +259024,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -260017,7 +260017,7 @@
     "min_vram_gb": 4.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -260025,13 +260025,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 24576,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -260615,7 +260615,7 @@
     "min_vram_gb": 4.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -260625,13 +260625,13 @@
     "hf_downloads": 89,
     "hf_likes": 1,
     "release_date": "2026-08-06",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -261177,7 +261177,7 @@
     "min_vram_gb": 4.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -261185,13 +261185,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3584,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -261205,7 +261205,7 @@
     "min_vram_gb": 4.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -261213,13 +261213,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3584,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -261263,7 +261263,7 @@
     "min_vram_gb": 4.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -261271,13 +261271,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3584,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -263626,7 +263626,7 @@
     "min_vram_gb": 4.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -263637,13 +263637,13 @@
     "hf_downloads": 33,
     "hf_likes": 2,
     "release_date": "2026-07-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -263690,7 +263690,7 @@
     "min_vram_gb": 4.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -263704,13 +263704,13 @@
     "hf_downloads": 116,
     "hf_likes": 1,
     "release_date": "2026-07-30",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -263724,7 +263724,7 @@
     "min_vram_gb": 4.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -263734,13 +263734,13 @@
     "hf_downloads": 345,
     "hf_likes": 12,
     "release_date": "2026-07-30",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -263850,7 +263850,7 @@
     "min_vram_gb": 4.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -263858,13 +263858,13 @@
     "hf_downloads": 25324,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -264064,7 +264064,7 @@
     "min_vram_gb": 4.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -264074,13 +264074,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -266196,7 +266196,7 @@
     "min_vram_gb": 4.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -266207,13 +266207,13 @@
     "hf_downloads": 0,
     "hf_likes": 2,
     "release_date": "2026-08-16",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -266972,7 +266972,7 @@
     "min_vram_gb": 4.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -266980,13 +266980,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -267090,7 +267090,7 @@
     "min_vram_gb": 4.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -267101,13 +267101,13 @@
     "hf_downloads": 5652,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -267122,7 +267122,7 @@
     "min_vram_gb": 4.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -267137,13 +267137,13 @@
     "hf_downloads": 452,
     "hf_likes": 11,
     "release_date": "2026-07-31",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -268822,7 +268822,7 @@
     "min_vram_gb": 5.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -268830,14 +268830,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -268855,7 +268855,7 @@
     "min_vram_gb": 5.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -268863,14 +268863,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -269352,7 +269352,7 @@
     "min_vram_gb": 5.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -269360,14 +269360,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
+    "shared_expert_intermediate_size": 1024,
     "_discovered": true,
     "license": "openmdw-1.1"
   },
@@ -269437,7 +269437,7 @@
     "min_vram_gb": 5.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -269445,13 +269445,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 46,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3584,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -269611,7 +269611,7 @@
     "min_vram_gb": 5.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -269619,13 +269619,13 @@
     "hf_downloads": 86,
     "hf_likes": 36,
     "release_date": "2024-07-17",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3584,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -269639,7 +269639,7 @@
     "min_vram_gb": 5.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -269647,13 +269647,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3584,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -269668,7 +269668,7 @@
     "min_vram_gb": 5.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -269676,13 +269676,13 @@
     "hf_downloads": 107,
     "hf_likes": 9,
     "release_date": "2024-09-26",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3584,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -269696,7 +269696,7 @@
     "min_vram_gb": 5.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -269704,13 +269704,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3584,
+    "vocab_size": 256000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -270705,7 +270705,7 @@
     "min_vram_gb": 5.4,
     "quantization": "AWQ-4bit",
     "format": "awq",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -270713,13 +270713,13 @@
     "hf_downloads": 22850,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -270855,7 +270855,7 @@
     "min_vram_gb": 5.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -270863,13 +270863,13 @@
     "hf_downloads": 76,
     "hf_likes": 6,
     "release_date": "2025-09-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 45,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 102400,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "gguf_sources": [
@@ -272348,7 +272348,7 @@
     "min_vram_gb": 5.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -272356,13 +272356,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 50,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32128,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "gguf_sources": [
@@ -273483,7 +273483,7 @@
     "min_vram_gb": 6.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -273491,13 +273491,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3840,
+    "vocab_size": 262208,
+    "moe_intermediate_size": 15360,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma",
@@ -273984,7 +273984,7 @@
     "min_vram_gb": 6.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -273992,13 +273992,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3840,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 15360,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other",
@@ -274019,7 +274019,7 @@
     "min_vram_gb": 6.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -274027,13 +274027,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3840,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 15360,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -274670,7 +274670,7 @@
     "min_vram_gb": 6.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -274678,13 +274678,13 @@
     "hf_downloads": 17226,
     "hf_likes": 11,
     "release_date": "2026-07-11",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3840,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 15360,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -274705,7 +274705,7 @@
     "min_vram_gb": 6.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -274713,13 +274713,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3840,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 15360,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -274779,7 +274779,7 @@
     "min_vram_gb": 6.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -274787,13 +274787,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3840,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 15360,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -275784,7 +275784,7 @@
     "min_vram_gb": 6.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -275792,14 +275792,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -275890,7 +275890,7 @@
     "min_vram_gb": 6.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 10485760,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -275898,13 +275898,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 202048,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama4",
@@ -276658,7 +276658,7 @@
     "min_vram_gb": 6.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -276668,13 +276668,13 @@
     "hf_downloads": 348,
     "hf_likes": 94,
     "release_date": "2025-11-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3840,
+    "vocab_size": 262208,
+    "moe_intermediate_size": 15360,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma",
@@ -277704,7 +277704,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -277712,13 +277712,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -277732,7 +277732,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -277740,13 +277740,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -277760,7 +277760,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -277768,13 +277768,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -277847,7 +277847,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -277855,13 +277855,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -277876,7 +277876,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024000,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -277886,13 +277886,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -277907,7 +277907,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -277915,13 +277915,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -277936,7 +277936,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -277944,13 +277944,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -277964,7 +277964,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -277972,13 +277972,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -278021,7 +278021,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -278029,13 +278029,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -278049,7 +278049,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -278057,13 +278057,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -279127,7 +279127,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -279135,13 +279135,13 @@
     "hf_downloads": 1993,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131074,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -279492,7 +279492,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -279500,13 +279500,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131075,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -279521,7 +279521,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -279529,13 +279529,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131075,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -279806,7 +279806,7 @@
     "min_vram_gb": 6.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -279814,13 +279814,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131078,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -280361,7 +280361,7 @@
     "min_vram_gb": 6.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1024000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -280369,13 +280369,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151081,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -280847,7 +280847,7 @@
     "min_vram_gb": 6.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -280855,13 +280855,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 219520,
+    "moe_intermediate_size": 16384,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -281435,7 +281435,7 @@
     "min_vram_gb": 6.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -281443,13 +281443,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 40,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 13696,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -281463,7 +281463,7 @@
     "min_vram_gb": 6.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -281471,13 +281471,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 3840,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 15360,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -283122,7 +283122,7 @@
     "min_vram_gb": 6.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16384,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -283130,13 +283130,13 @@
     "hf_downloads": 2560,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 40,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 32016,
+    "moe_intermediate_size": 13824,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama2"
@@ -283654,7 +283654,7 @@
     "min_vram_gb": 6.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -283662,14 +283662,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
+    "shared_expert_intermediate_size": 1024,
     "_discovered": true,
     "license": "openmdw-1.1"
   },
@@ -284358,7 +284358,7 @@
     "min_vram_gb": 6.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2097152,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -284366,12 +284366,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -284631,13 +284631,13 @@
     "hf_downloads": 312,
     "hf_likes": 28,
     "release_date": "2026-04-29",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -284660,13 +284660,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -284692,13 +284692,13 @@
     "hf_downloads": 206,
     "hf_likes": 5,
     "release_date": "2026-05-05",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -284721,13 +284721,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 16,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -284800,7 +284800,7 @@
     "min_vram_gb": 7.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -284808,13 +284808,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.1",
@@ -284963,7 +284963,7 @@
     "min_vram_gb": 7.0,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -284973,14 +284973,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -286221,7 +286221,7 @@
     "min_vram_gb": 7.4,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -286229,14 +286229,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -286463,7 +286463,7 @@
     "min_vram_gb": 7.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16384,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -286475,13 +286475,13 @@
     "hf_downloads": 10,
     "hf_likes": 3,
     "release_date": "2026-08-07",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 10,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 17920,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -286871,7 +286871,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -286879,13 +286879,13 @@
     "hf_downloads": 1191,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151665,
+    "moe_intermediate_size": 13824,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -286939,7 +286939,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -286947,13 +286947,13 @@
     "hf_downloads": 1113,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151665,
+    "moe_intermediate_size": 13824,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -286974,7 +286974,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -286982,13 +286982,13 @@
     "hf_downloads": 1585,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151665,
+    "moe_intermediate_size": 13824,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -287555,7 +287555,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -287563,13 +287563,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -287584,7 +287584,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -287592,13 +287592,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -287613,7 +287613,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -287623,13 +287623,13 @@
     "hf_downloads": 5258,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -287679,7 +287679,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 163840,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -287687,13 +287687,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -287739,7 +287739,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -287749,13 +287749,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -287770,7 +287770,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -287780,13 +287780,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -287801,7 +287801,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -287809,13 +287809,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -287830,7 +287830,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -287840,13 +287840,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -287860,7 +287860,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -287868,13 +287868,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -287888,7 +287888,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -287896,13 +287896,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -287917,7 +287917,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -287929,13 +287929,13 @@
     "hf_downloads": 713,
     "hf_likes": 11,
     "release_date": "2025-10-23",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -289788,7 +289788,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -289796,13 +289796,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 13824,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -289817,7 +289817,7 @@
     "min_vram_gb": 7.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -289825,13 +289825,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 13824,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-sa-4.0"
@@ -291117,7 +291117,7 @@
     "min_vram_gb": 7.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 65536,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -291125,13 +291125,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 50,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -291271,7 +291271,7 @@
     "min_vram_gb": 7.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -291279,13 +291279,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "gguf_sources": [
@@ -292588,7 +292588,7 @@
     "min_vram_gb": 7.9,
     "quantization": "AutoRound-4bit",
     "format": "autoround",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -292598,13 +292598,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 96,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -292740,7 +292740,7 @@
     "min_vram_gb": 8.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -292748,13 +292748,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 52,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 2688,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 1856,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "openmdw-1.1"
@@ -292909,7 +292909,7 @@
     "min_vram_gb": 8.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -292921,13 +292921,13 @@
     "hf_downloads": 362,
     "hf_likes": 6,
     "release_date": "2026-07-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -293501,7 +293501,7 @@
     "min_vram_gb": 8.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 163840,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -293509,13 +293509,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 27,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 128896,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "researchrail",
@@ -293534,7 +293534,7 @@
     "min_vram_gb": 8.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1310720,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -293542,13 +293542,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 27,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 128896,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "researchrail",
@@ -293567,7 +293567,7 @@
     "min_vram_gb": 8.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2621440,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -293575,13 +293575,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 27,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 128896,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "researchrail",
@@ -293600,7 +293600,7 @@
     "min_vram_gb": 8.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1310720,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -293608,13 +293608,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 27,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 128896,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "researchrail",
@@ -293633,7 +293633,7 @@
     "min_vram_gb": 8.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 163840,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -293641,13 +293641,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 27,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 128896,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "researchrail",
@@ -294142,7 +294142,7 @@
     "min_vram_gb": 8.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -294150,13 +294150,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151552,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -294950,7 +294950,7 @@
     "min_vram_gb": 8.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -294958,13 +294958,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 12288,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -295614,7 +295614,7 @@
     "min_vram_gb": 8.6,
     "quantization": "AutoRound-4bit",
     "format": "autoround",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -295628,13 +295628,13 @@
     "hf_downloads": 223,
     "hf_likes": 2,
     "release_date": "2026-08-14",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -295786,7 +295786,7 @@
     "min_vram_gb": 8.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -295794,13 +295794,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 126464,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -296925,7 +296925,7 @@
     "min_vram_gb": 9.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -296933,13 +296933,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32001,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-nd-4.0"
@@ -297647,7 +297647,7 @@
     "min_vram_gb": 9.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -297657,13 +297657,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "is_moe": true,
@@ -299103,7 +299103,7 @@
     "min_vram_gb": 10.0,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -299111,13 +299111,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 157184,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -299477,7 +299477,7 @@
     "min_vram_gb": 10.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -299488,13 +299488,13 @@
     "hf_downloads": 0,
     "hf_likes": 2,
     "release_date": "2026-08-16",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 44,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -299923,7 +299923,7 @@
     "num_key_value_heads": null,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 49152,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -300599,7 +300599,7 @@
     "min_vram_gb": 10.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -300607,14 +300607,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
+    "shared_expert_intermediate_size": 1024,
     "_discovered": true,
     "license": "openmdw-1.1"
   },
@@ -300659,7 +300659,7 @@
     "min_vram_gb": 10.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -300667,13 +300667,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 1792,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -301654,7 +301654,7 @@
     "min_vram_gb": 10.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -301662,13 +301662,13 @@
     "hf_downloads": 5125,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2880,
+    "vocab_size": 201088,
+    "moe_intermediate_size": 2880,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -301747,7 +301747,7 @@
     "min_vram_gb": 10.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "languages": [
@@ -301760,13 +301760,13 @@
     "hf_downloads": 204,
     "hf_likes": 3,
     "release_date": "2026-08-07",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2880,
+    "vocab_size": 201088,
+    "moe_intermediate_size": 2880,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -301781,7 +301781,7 @@
     "min_vram_gb": 10.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -301789,13 +301789,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2880,
+    "vocab_size": 201088,
+    "moe_intermediate_size": 2880,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -301810,7 +301810,7 @@
     "min_vram_gb": 10.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -301818,13 +301818,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2880,
+    "vocab_size": 201088,
+    "moe_intermediate_size": 2880,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-4.0"
@@ -301839,7 +301839,7 @@
     "min_vram_gb": 10.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -301847,13 +301847,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2880,
+    "vocab_size": 201088,
+    "moe_intermediate_size": 2880,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -301896,7 +301896,7 @@
     "min_vram_gb": 10.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -301904,13 +301904,13 @@
     "hf_downloads": 145337,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2880,
+    "vocab_size": 201088,
+    "moe_intermediate_size": 2880,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -303108,7 +303108,7 @@
     "min_vram_gb": 10.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -303118,14 +303118,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -303544,7 +303544,7 @@
     "min_vram_gb": 10.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -303552,13 +303552,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 2816,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 704,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -303665,7 +303665,7 @@
     "min_vram_gb": 11.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16384,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -303673,12 +303673,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 52,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -304382,7 +304382,7 @@
     "min_vram_gb": 11.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -304392,13 +304392,13 @@
     "hf_downloads": 12644,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 88,
+    "num_attention_heads": 96,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 12288,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -304413,7 +304413,7 @@
     "min_vram_gb": 11.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -304421,13 +304421,13 @@
     "hf_downloads": 4617,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 24,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2880,
+    "vocab_size": 201088,
+    "moe_intermediate_size": 2880,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -304631,7 +304631,7 @@
     "min_vram_gb": 11.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -304639,13 +304639,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 56,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 32768,
+    "moe_intermediate_size": 16384,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -304660,7 +304660,7 @@
     "min_vram_gb": 11.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -304668,13 +304668,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 56,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 32768,
+    "moe_intermediate_size": 16384,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -304689,7 +304689,7 @@
     "min_vram_gb": 11.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -304697,13 +304697,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 56,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 32768,
+    "moe_intermediate_size": 16384,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mrl"
@@ -305402,7 +305402,7 @@
     "min_vram_gb": 11.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 202752,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -305410,13 +305410,13 @@
     "hf_downloads": 1338,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 47,
+    "num_attention_heads": 20,
+    "num_key_value_heads": 20,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -305441,7 +305441,7 @@
     "min_vram_gb": 11.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -305449,13 +305449,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 36864,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -305470,7 +305470,7 @@
     "min_vram_gb": 11.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -305478,13 +305478,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 128000,
+    "moe_intermediate_size": 36864,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -306377,7 +306377,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -306385,13 +306385,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -306412,7 +306412,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -306420,13 +306420,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -306447,7 +306447,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -306455,13 +306455,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other",
@@ -306516,7 +306516,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 65280,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -306524,13 +306524,13 @@
     "hf_downloads": 2007,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -306551,7 +306551,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -306559,13 +306559,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -306580,7 +306580,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -306588,13 +306588,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other",
@@ -306615,7 +306615,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -306623,13 +306623,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -306650,7 +306650,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -306658,13 +306658,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "gguf_sources": [
@@ -307603,7 +307603,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -307611,13 +307611,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131074,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -307856,7 +307856,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -307864,13 +307864,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131076,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "gguf_sources": [
@@ -308110,7 +308110,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -308120,13 +308120,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131078,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -308182,7 +308182,7 @@
     "min_vram_gb": 12.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -308190,13 +308190,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131084,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -308545,7 +308545,7 @@
     "min_vram_gb": 12.2,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -308553,13 +308553,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -308574,7 +308574,7 @@
     "min_vram_gb": 12.2,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -308582,13 +308582,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -308603,7 +308603,7 @@
     "min_vram_gb": 12.2,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -308611,13 +308611,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -308632,7 +308632,7 @@
     "min_vram_gb": 12.2,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -308640,13 +308640,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -308661,7 +308661,7 @@
     "min_vram_gb": 12.2,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -308669,13 +308669,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2048,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "lfm1.0"
@@ -309194,7 +309194,7 @@
     "min_vram_gb": 12.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -309202,14 +309202,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
+    "shared_expert_intermediate_size": 1024,
     "_discovered": true,
     "license": "openmdw-1.1"
   },
@@ -309254,7 +309254,7 @@
     "min_vram_gb": 12.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -309262,13 +309262,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32000,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "is_moe": true,
@@ -314183,7 +314183,7 @@
     "min_vram_gb": 12.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -314191,13 +314191,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 120832,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -314649,7 +314649,7 @@
     "min_vram_gb": 12.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -314657,13 +314657,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -314682,7 +314682,7 @@
     "min_vram_gb": 12.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -314690,13 +314690,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 200064,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -315245,7 +315245,7 @@
     "min_vram_gb": 12.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -315253,13 +315253,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 52,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 2688,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 1856,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "openmdw-1.1"
@@ -316803,7 +316803,7 @@
     "min_vram_gb": 13.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -316813,13 +316813,13 @@
     "hf_downloads": 356,
     "hf_likes": 41,
     "release_date": "2026-06-18",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 2816,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 704,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -316908,7 +316908,7 @@
     "min_vram_gb": 13.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -316919,13 +316919,13 @@
     "hf_downloads": 17427,
     "hf_likes": 59,
     "release_date": "2026-04-03",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 2816,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 704,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -317428,7 +317428,7 @@
     "min_vram_gb": 13.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -317438,13 +317438,13 @@
     "hf_downloads": 0,
     "hf_likes": 2,
     "release_date": "2026-08-16",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 2816,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 704,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -317459,7 +317459,7 @@
     "min_vram_gb": 13.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -317469,13 +317469,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 2816,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 704,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -318711,7 +318711,7 @@
     "min_vram_gb": 13.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -318721,13 +318721,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 55,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -318921,7 +318921,7 @@
     "min_vram_gb": 13.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -318929,13 +318929,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 2816,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 704,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -319045,7 +319045,7 @@
     "min_vram_gb": 13.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -319053,13 +319053,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 30,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 8,
+    "head_dim": 256,
+    "hidden_size": 2816,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 704,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -319700,7 +319700,7 @@
     "min_vram_gb": 13.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -319710,14 +319710,14 @@
     "hf_downloads": 606,
     "hf_likes": 6,
     "release_date": "2026-07-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -319741,7 +319741,7 @@
     "min_vram_gb": 13.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -319751,13 +319751,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 55,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 9728,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -319869,7 +319869,7 @@
     "min_vram_gb": 13.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -319879,13 +319879,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -319900,7 +319900,7 @@
     "min_vram_gb": 13.8,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -319910,13 +319910,13 @@
     "hf_downloads": 7966,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -320024,7 +320024,7 @@
     "min_vram_gb": 13.8,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -320032,13 +320032,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -321604,7 +321604,7 @@
     "min_vram_gb": 13.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -321612,13 +321612,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -322351,7 +322351,7 @@
     "min_vram_gb": 13.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -322362,13 +322362,13 @@
     "hf_downloads": 77,
     "hf_likes": 1,
     "release_date": "2026-08-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -322565,7 +322565,7 @@
     "min_vram_gb": 13.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -322573,13 +322573,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 62,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 128,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "health-ai-developer-foundations"
@@ -323145,7 +323145,7 @@
     "min_vram_gb": 14.0,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -323155,13 +323155,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -323176,7 +323176,7 @@
     "min_vram_gb": 14.0,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -323186,13 +323186,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -323852,7 +323852,7 @@
     "min_vram_gb": 14.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -323860,13 +323860,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -324463,7 +324463,7 @@
     "min_vram_gb": 14.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -324474,13 +324474,13 @@
     "hf_downloads": 315,
     "hf_likes": 3,
     "release_date": "2026-08-14",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -324527,7 +324527,7 @@
     "min_vram_gb": 14.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -324538,13 +324538,13 @@
     "hf_downloads": 45,
     "hf_likes": 2,
     "release_date": "2026-08-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -324559,7 +324559,7 @@
     "min_vram_gb": 14.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -324570,13 +324570,13 @@
     "hf_downloads": 18,
     "hf_likes": 1,
     "release_date": "2026-08-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -324941,7 +324941,7 @@
     "min_vram_gb": 14.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -324949,13 +324949,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -328020,11 +328020,11 @@
     "hf_downloads": 845,
     "hf_likes": 7,
     "release_date": "2026-07-16",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 24,
+    "head_dim": 213,
+    "hidden_size": 5120,
     "vocab_size": null,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
@@ -330024,7 +330024,7 @@
     "min_vram_gb": 14.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [
       "vision"
@@ -330037,13 +330037,13 @@
     "hf_downloads": 749,
     "hf_likes": 26,
     "release_date": "2026-06-16",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -330420,7 +330420,7 @@
     "min_vram_gb": 14.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -330430,13 +330430,13 @@
     "hf_downloads": 754,
     "hf_likes": 36,
     "release_date": "2026-07-07",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -330451,7 +330451,7 @@
     "min_vram_gb": 14.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -330465,13 +330465,13 @@
     "hf_downloads": 201795,
     "hf_likes": 2,
     "release_date": "2026-07-17",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -330492,7 +330492,7 @@
     "min_vram_gb": 14.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -330503,13 +330503,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -330530,7 +330530,7 @@
     "min_vram_gb": 14.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -330540,13 +330540,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -330599,7 +330599,7 @@
     "min_vram_gb": 14.2,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -330610,13 +330610,13 @@
     "hf_downloads": 332,
     "hf_likes": 1,
     "release_date": "2026-08-14",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -330701,7 +330701,7 @@
     "min_vram_gb": 14.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -330716,13 +330716,13 @@
     "hf_downloads": 705,
     "hf_likes": 1,
     "release_date": "2026-08-15",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -331244,7 +331244,7 @@
     "min_vram_gb": 14.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -331252,13 +331252,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 24,
+    "num_key_value_heads": 4,
+    "head_dim": 256,
+    "hidden_size": 5120,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 17408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "qwen",
@@ -334088,7 +334088,7 @@
     "min_vram_gb": 14.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -334098,14 +334098,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -334953,7 +334953,7 @@
     "min_vram_gb": 15.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -334963,13 +334963,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 94,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -336435,7 +336435,7 @@
     "min_vram_gb": 15.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 500000,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -336443,13 +336443,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 49,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -336571,7 +336571,7 @@
     "min_vram_gb": 15.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1010000,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -336581,13 +336581,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151669,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -337886,7 +337886,7 @@
     "min_vram_gb": 15.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -337894,13 +337894,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0",
@@ -337998,7 +337998,7 @@
     "min_vram_gb": 15.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -338008,13 +338008,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "is_moe": true,
@@ -338073,7 +338073,7 @@
     "min_vram_gb": 15.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -338083,13 +338083,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -338108,7 +338108,7 @@
     "min_vram_gb": 15.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -338118,14 +338118,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
+    "shared_expert_intermediate_size": 0,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -338143,7 +338143,7 @@
     "min_vram_gb": 15.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -338153,14 +338153,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
+    "shared_expert_intermediate_size": 0,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -338178,7 +338178,7 @@
     "min_vram_gb": 15.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -338186,13 +338186,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -338211,7 +338211,7 @@
     "min_vram_gb": 15.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -338219,14 +338219,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
+    "shared_expert_intermediate_size": 0,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -338279,7 +338279,7 @@
     "min_vram_gb": 15.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -338289,13 +338289,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -340053,7 +340053,7 @@
     "min_vram_gb": 15.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -340061,13 +340061,13 @@
     "hf_downloads": 3662,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -340337,7 +340337,7 @@
     "min_vram_gb": 15.8,
     "quantization": "AWQ-4bit",
     "format": "awq",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -340347,14 +340347,14 @@
     "hf_downloads": 1565,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
+    "shared_expert_intermediate_size": 0,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -341103,7 +341103,7 @@
     "min_vram_gb": 16.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 202752,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -341111,13 +341111,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 47,
+    "num_attention_heads": 20,
+    "num_key_value_heads": 20,
+    "head_dim": 102,
+    "hidden_size": 2048,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -341495,7 +341495,7 @@
     "min_vram_gb": 16.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -341505,13 +341505,13 @@
     "hf_downloads": 267,
     "hf_likes": 63,
     "release_date": "2026-04-03",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -341526,7 +341526,7 @@
     "min_vram_gb": 16.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -341539,13 +341539,13 @@
     "hf_downloads": 460,
     "hf_likes": 9,
     "release_date": "2026-04-04",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -341560,7 +341560,7 @@
     "min_vram_gb": 16.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -341571,13 +341571,13 @@
     "hf_downloads": 220,
     "hf_likes": 28,
     "release_date": "2026-04-10",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -341592,7 +341592,7 @@
     "min_vram_gb": 16.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -341602,13 +341602,13 @@
     "hf_downloads": 202,
     "hf_likes": 17,
     "release_date": "2026-05-13",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -341654,7 +341654,7 @@
     "min_vram_gb": 16.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -341662,13 +341662,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -342489,7 +342489,7 @@
     "min_vram_gb": 16.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -342497,13 +342497,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -342611,7 +342611,7 @@
     "min_vram_gb": 16.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -342621,13 +342621,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "openmdw-1.1"
@@ -342642,7 +342642,7 @@
     "min_vram_gb": 16.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -342650,13 +342650,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -342671,7 +342671,7 @@
     "min_vram_gb": 16.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -342679,13 +342679,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -342700,7 +342700,7 @@
     "min_vram_gb": 16.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -342708,13 +342708,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -343133,7 +343133,7 @@
     "min_vram_gb": 16.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -343141,13 +343141,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -343166,7 +343166,7 @@
     "min_vram_gb": 16.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -343174,13 +343174,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -344626,7 +344626,7 @@
     "min_vram_gb": 16.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -344634,13 +344634,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 52,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 2688,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 1856,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "openmdw-1.1"
@@ -344655,7 +344655,7 @@
     "min_vram_gb": 16.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -344664,12 +344664,12 @@
     "hf_likes": 0,
     "release_date": null,
     "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 2688,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 1856,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "nvidia-nemotron-open-model-license"
@@ -345161,7 +345161,7 @@
     "min_vram_gb": 16.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -345169,13 +345169,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 52,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 2688,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 1856,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "nvidia-open-model-agreement"
@@ -345345,7 +345345,7 @@
     "min_vram_gb": 32.5,
     "quantization": "AWQ-8bit",
     "format": "awq",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -345353,13 +345353,13 @@
     "hf_downloads": 52,
     "hf_likes": 3,
     "release_date": "2025-11-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 46,
+    "num_attention_heads": 96,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151552,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -345676,7 +345676,7 @@
     "min_vram_gb": 16.3,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -345684,13 +345684,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -347272,7 +347272,7 @@
     "min_vram_gb": 16.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -347280,13 +347280,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 61,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 151345,
+    "moe_intermediate_size": 23040,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -347747,7 +347747,7 @@
     "min_vram_gb": 16.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -347757,13 +347757,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -347778,7 +347778,7 @@
     "min_vram_gb": 16.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -347786,13 +347786,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -347807,7 +347807,7 @@
     "min_vram_gb": 16.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -347815,13 +347815,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "gemma"
@@ -348171,7 +348171,7 @@
     "min_vram_gb": 16.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -348179,13 +348179,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 16,
+    "head_dim": 256,
+    "hidden_size": 5376,
+    "vocab_size": 262144,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -348319,7 +348319,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -348327,13 +348327,13 @@
     "hf_downloads": 25687,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151689,
+    "moe_intermediate_size": 25600,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-sa-4.0"
@@ -348406,7 +348406,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -348414,13 +348414,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151665,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -348434,7 +348434,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -348442,13 +348442,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151665,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -348937,7 +348937,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -348945,13 +348945,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 25600,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -349003,7 +349003,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -349011,13 +349011,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 25600,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -349032,7 +349032,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -349042,13 +349042,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 25600,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "nvidia-open-model-license"
@@ -349063,7 +349063,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -349071,13 +349071,13 @@
     "hf_downloads": 4896,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 25600,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -350228,7 +350228,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -350236,13 +350236,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -350315,7 +350315,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -350323,13 +350323,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -350344,7 +350344,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -350352,13 +350352,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -350373,7 +350373,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 65536,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -350381,13 +350381,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -350430,7 +350430,7 @@
     "min_vram_gb": 16.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -350438,13 +350438,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.1"
@@ -351549,7 +351549,7 @@
     "min_vram_gb": 16.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -351557,13 +351557,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 52,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 2688,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 1856,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "openmdw-1.1",
@@ -352121,7 +352121,7 @@
     "min_vram_gb": 17.1,
     "quantization": "AWQ-4bit",
     "format": "awq",
-    "context_length": 4096,
+    "context_length": 65536,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -352129,13 +352129,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 62,
+    "num_attention_heads": 56,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 7168,
+    "vocab_size": 32256,
+    "moe_intermediate_size": 19200,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "deepseek"
@@ -352996,7 +352996,7 @@
     "min_vram_gb": 17.1,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -353004,14 +353004,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "openmdw-1.1"
   },
@@ -353025,7 +353025,7 @@
     "min_vram_gb": 17.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -353033,14 +353033,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "openmdw-1.1"
   },
@@ -353119,7 +353119,7 @@
     "min_vram_gb": 17.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -353129,14 +353129,14 @@
     "hf_downloads": 92,
     "hf_likes": 2,
     "release_date": "2026-08-03",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "openmdw-1.1-and-kimi-modified-mit"
   },
@@ -353293,7 +353293,7 @@
     "min_vram_gb": 17.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -353303,13 +353303,13 @@
     "hf_downloads": 2235,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -353367,7 +353367,7 @@
     "min_vram_gb": 17.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -353375,13 +353375,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -353402,7 +353402,7 @@
     "min_vram_gb": 17.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 128000,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -353410,13 +353410,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -357261,7 +357261,7 @@
     "min_vram_gb": 17.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -357269,13 +357269,13 @@
     "hf_downloads": 321,
     "hf_likes": 19,
     "release_date": "2023-10-12",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 143973,
+    "moe_intermediate_size": 24576,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -358387,7 +358387,7 @@
     "min_vram_gb": 17.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -358395,13 +358395,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 196608,
+    "moe_intermediate_size": 1280,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "upstage-solar-license"
@@ -358693,7 +358693,7 @@
     "min_vram_gb": 17.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -358701,14 +358701,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -358726,7 +358726,7 @@
     "min_vram_gb": 17.8,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -358736,14 +358736,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -358796,7 +358796,7 @@
     "min_vram_gb": 17.8,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -358809,14 +358809,14 @@
     "hf_downloads": 155,
     "hf_likes": 7,
     "release_date": "2026-06-17",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -360743,7 +360743,7 @@
     "min_vram_gb": 17.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -360751,14 +360751,14 @@
     "hf_downloads": 2791,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "gguf_sources": [
@@ -360778,7 +360778,7 @@
     "min_vram_gb": 17.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -360786,14 +360786,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0"
   },
@@ -360807,7 +360807,7 @@
     "min_vram_gb": 17.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -360815,14 +360815,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "gguf_sources": [
@@ -361432,7 +361432,7 @@
     "min_vram_gb": 17.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -361440,13 +361440,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 64,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -362118,7 +362118,7 @@
     "min_vram_gb": 18.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -362126,14 +362126,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "is_moe": true,
     "num_experts": 256,
@@ -362150,7 +362150,7 @@
     "min_vram_gb": 18.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -362158,14 +362158,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -362322,7 +362322,7 @@
     "min_vram_gb": 18.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -362333,14 +362333,14 @@
     "hf_downloads": 72999,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -362358,7 +362358,7 @@
     "min_vram_gb": 18.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -362371,14 +362371,14 @@
     "hf_downloads": 574,
     "hf_likes": 372,
     "release_date": "2026-03-23",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -362396,7 +362396,7 @@
     "min_vram_gb": 18.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -362407,14 +362407,14 @@
     "hf_downloads": 2613,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -362538,7 +362538,7 @@
     "min_vram_gb": 18.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -362546,14 +362546,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -362571,7 +362571,7 @@
     "min_vram_gb": 18.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -362579,14 +362579,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -362604,7 +362604,7 @@
     "min_vram_gb": 18.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -362612,14 +362612,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "mit",
     "is_moe": true,
@@ -364175,7 +364175,7 @@
     "min_vram_gb": 18.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -364183,13 +364183,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 88,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 2688,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -365235,7 +365235,7 @@
     "min_vram_gb": 18.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -365243,14 +365243,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "agpl-3.0",
     "is_moe": true,
@@ -365274,7 +365274,7 @@
     "min_vram_gb": 18.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -365282,14 +365282,14 @@
     "hf_downloads": 13229,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -365307,7 +365307,7 @@
     "min_vram_gb": 18.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -365315,14 +365315,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -365340,7 +365340,7 @@
     "min_vram_gb": 18.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -365348,14 +365348,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "mit",
     "is_moe": true,
@@ -365474,7 +365474,7 @@
     "min_vram_gb": 18.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision",
@@ -365485,14 +365485,14 @@
     "hf_downloads": 1296,
     "hf_likes": 353,
     "release_date": "2026-02-27",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -365581,7 +365581,7 @@
     "min_vram_gb": 18.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [
       "tool_use"
@@ -365591,14 +365591,14 @@
     "hf_downloads": 2011,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -365649,7 +365649,7 @@
     "min_vram_gb": 18.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -365657,14 +365657,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 40,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "agpl-3.0",
     "is_moe": true,
@@ -367780,7 +367780,7 @@
     "min_vram_gb": 19.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -367788,13 +367788,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 47,
+    "num_attention_heads": 20,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 103424,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -369085,7 +369085,7 @@
     "min_vram_gb": 20.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -369093,13 +369093,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 27648,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "kwaipilot-license"
@@ -369568,7 +369568,7 @@
     "min_vram_gb": 21.2,
     "quantization": "AutoRound-4bit",
     "format": "autoround",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -369576,13 +369576,13 @@
     "hf_downloads": 26243,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -369693,13 +369693,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32064,
+    "moe_intermediate_size": 6400,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -369744,7 +369744,7 @@
     "min_vram_gb": 21.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1010000,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -369754,13 +369754,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 67,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151669,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -369814,7 +369814,7 @@
     "min_vram_gb": 21.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -369824,14 +369824,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 67,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
+    "shared_expert_intermediate_size": 0,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -369849,7 +369849,7 @@
     "min_vram_gb": 21.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -369859,13 +369859,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 67,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -369884,7 +369884,7 @@
     "min_vram_gb": 21.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -369900,13 +369900,13 @@
     "hf_downloads": 66,
     "hf_likes": 7,
     "release_date": "2025-07-30",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 67,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -369999,7 +369999,7 @@
     "min_vram_gb": 22.2,
     "quantization": "AutoRound-4bit",
     "format": "autoround",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -370007,13 +370007,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -370032,7 +370032,7 @@
     "min_vram_gb": 22.2,
     "quantization": "AutoRound-4bit",
     "format": "autoround",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -370040,13 +370040,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -370365,7 +370365,7 @@
     "min_vram_gb": 23.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -370373,13 +370373,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2880,
+    "vocab_size": 201088,
+    "moe_intermediate_size": 2880,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -370422,7 +370422,7 @@
     "min_vram_gb": 23.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -370432,13 +370432,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 79,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -370453,7 +370453,7 @@
     "min_vram_gb": 23.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -370463,13 +370463,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 79,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -371127,7 +371127,7 @@
     "min_vram_gb": 23.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -371135,13 +371135,13 @@
     "hf_downloads": 1598,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 32002,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -371344,7 +371344,7 @@
     "min_vram_gb": 24.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -371352,13 +371352,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 120832,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -371638,13 +371638,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 27,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 72,
+    "hidden_size": 2304,
+    "vocab_size": 163840,
+    "moe_intermediate_size": 1024,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -371702,13 +371702,13 @@
     "hf_downloads": 1216,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 27,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 72,
+    "hidden_size": 2304,
+    "vocab_size": 163840,
+    "moe_intermediate_size": 1024,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -372284,7 +372284,7 @@
     "min_vram_gb": 26.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -372292,13 +372292,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 65536,
+    "moe_intermediate_size": 14336,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -372525,7 +372525,7 @@
     "min_vram_gb": 27.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -372535,13 +372535,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 84,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -373001,7 +373001,7 @@
     "min_vram_gb": 30.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -373009,13 +373009,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 196608,
+    "moe_intermediate_size": 1280,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "upstage-solar-license"
@@ -373549,7 +373549,7 @@
     "min_vram_gb": 32.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 167772160,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -373557,13 +373557,13 @@
     "hf_downloads": 1025,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 202048,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.1",
@@ -373907,7 +373907,7 @@
     "min_vram_gb": 34.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -373915,13 +373915,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 157184,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -374348,7 +374348,7 @@
     "min_vram_gb": 35.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 327680,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -374356,12 +374356,12 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 64,
+    "hidden_size": 3072,
+    "vocab_size": 131072,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -375294,13 +375294,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 32016,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama2",
@@ -375656,7 +375656,7 @@
     "min_vram_gb": 35.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 117964800,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -375665,11 +375665,11 @@
     "hf_likes": 0,
     "release_date": null,
     "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 96,
+    "hidden_size": 3072,
+    "vocab_size": 131072,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -376930,7 +376930,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -376940,13 +376940,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-4.0"
@@ -376961,7 +376961,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -376971,13 +376971,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -376991,7 +376991,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -376999,13 +376999,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -377019,7 +377019,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -377027,13 +377027,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.3"
@@ -377077,7 +377077,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -377088,13 +377088,13 @@
     "hf_downloads": 106,
     "hf_likes": 12,
     "release_date": "2026-01-31",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -377139,7 +377139,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -377147,13 +377147,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -377196,7 +377196,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 8192,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -377206,13 +377206,13 @@
     "hf_downloads": 2223,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3",
@@ -377237,7 +377237,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -377245,13 +377245,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -377265,7 +377265,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "tool_use"
@@ -377275,13 +377275,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.1"
@@ -377296,7 +377296,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -377304,13 +377304,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.1"
@@ -377325,7 +377325,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -377333,13 +377333,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.3"
@@ -377382,7 +377382,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Advanced reasoning, chain-of-thought",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -377390,13 +377390,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.3"
@@ -377411,7 +377411,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -377419,13 +377419,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.1"
@@ -377440,7 +377440,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -377448,13 +377448,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.3"
@@ -377469,7 +377469,7 @@
     "min_vram_gb": 36.1,
     "quantization": "AWQ-4bit",
     "format": "awq",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -377477,13 +377477,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -377526,7 +377526,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -377534,13 +377534,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.3"
@@ -377645,7 +377645,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -377653,13 +377653,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -379799,7 +379799,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -379807,13 +379807,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128259,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.1"
@@ -380117,7 +380117,7 @@
     "min_vram_gb": 36.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -380127,13 +380127,13 @@
     "hf_downloads": 47134,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -380452,7 +380452,7 @@
     "min_vram_gb": 36.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 524288,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -380460,13 +380460,13 @@
     "hf_downloads": 16730,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 43008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -380808,7 +380808,7 @@
     "min_vram_gb": 36.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -380818,13 +380818,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 266752,
+    "moe_intermediate_size": 43008,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -381793,7 +381793,7 @@
     "min_vram_gb": 37.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -381801,13 +381801,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 29568,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "cc-by-nc-sa-4.0"
@@ -381880,7 +381880,7 @@
     "min_vram_gb": 37.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -381888,13 +381888,13 @@
     "hf_downloads": 12267,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 29568,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -381919,7 +381919,7 @@
     "min_vram_gb": 37.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -381929,13 +381929,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 29568,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "geogpt"
@@ -381950,7 +381950,7 @@
     "min_vram_gb": 37.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -381958,13 +381958,13 @@
     "hf_downloads": 3037,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 152064,
+    "moe_intermediate_size": 29568,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -382429,7 +382429,7 @@
     "min_vram_gb": 37.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -382437,13 +382437,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 157184,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -382890,7 +382890,7 @@
     "min_vram_gb": 38.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -382898,13 +382898,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 157184,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -383195,7 +383195,7 @@
     "min_vram_gb": 40.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -383203,13 +383203,13 @@
     "hf_downloads": 1300,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 88,
+    "num_attention_heads": 96,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 12288,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -384157,7 +384157,7 @@
     "min_vram_gb": 40.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -384167,14 +384167,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 2048,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 512,
+    "shared_expert_intermediate_size": 512,
     "_discovered": true,
     "license": "apache-2.0"
   },
@@ -385498,7 +385498,7 @@
     "min_vram_gb": 42.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -385506,13 +385506,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 46,
+    "num_attention_heads": 96,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151552,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -385655,7 +385655,7 @@
     "min_vram_gb": 42.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -385673,12 +385673,12 @@
     "hf_likes": 7,
     "release_date": "2026-07-13",
     "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 21504,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "openmdw-1.1"
@@ -385950,7 +385950,7 @@
     "min_vram_gb": 44.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 196608,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -385958,13 +385958,13 @@
     "hf_downloads": 14412,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 62,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 200064,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other",
@@ -386076,7 +386076,7 @@
     "min_vram_gb": 45.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Instruction following, chat",
     "capabilities": [
       "vision",
@@ -386087,13 +386087,13 @@
     "hf_downloads": 1066,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 100,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 28672,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3.2"
@@ -386268,7 +386268,7 @@
     "min_vram_gb": 47.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -386276,13 +386276,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -386399,7 +386399,7 @@
     "min_vram_gb": 49.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 32768,
     "use_case": "Instruction following, chat",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -386407,13 +386407,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 86,
+    "num_attention_heads": 72,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 9216,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 32768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -386563,7 +386563,7 @@
     "min_vram_gb": 50.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 196608,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -386571,13 +386571,13 @@
     "hf_downloads": 14145,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 62,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 200064,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -386625,7 +386625,7 @@
     "min_vram_gb": 50.7,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -386635,14 +386635,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 3072,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 1024,
+    "shared_expert_intermediate_size": 1024,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -386928,7 +386928,7 @@
     "min_vram_gb": 52.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -386936,13 +386936,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 196608,
+    "moe_intermediate_size": 1280,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "upstage-solar-license"
@@ -387019,7 +387019,7 @@
     "min_vram_gb": 52.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -387027,13 +387027,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 32,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 157184,
+    "moe_intermediate_size": 1024,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -387375,7 +387375,7 @@
     "min_vram_gb": 53.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -387383,13 +387383,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 120832,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -387738,7 +387738,7 @@
     "min_vram_gb": 54.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -387746,13 +387746,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 46,
+    "num_attention_heads": 96,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151552,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -387767,7 +387767,7 @@
     "min_vram_gb": 54.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -387779,13 +387779,13 @@
     "hf_downloads": 2765,
     "hf_likes": 11,
     "release_date": "2025-07-28",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 46,
+    "num_attention_heads": 96,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151552,
+    "moe_intermediate_size": 1408,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -388268,7 +388268,7 @@
     "min_vram_gb": 55.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 167772160,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -388278,13 +388278,13 @@
     "hf_downloads": 14342,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 40,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 5120,
+    "vocab_size": 202048,
+    "moe_intermediate_size": 8192,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "is_moe": true,
@@ -388913,7 +388913,7 @@
     "min_vram_gb": 58.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 196608,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -388921,13 +388921,13 @@
     "hf_downloads": 4538,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 62,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 200064,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other",
@@ -389004,7 +389004,7 @@
     "min_vram_gb": 59.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -389012,13 +389012,13 @@
     "hf_downloads": 10262,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 192,
+    "hidden_size": 6144,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -389576,7 +389576,7 @@
     "min_vram_gb": 59.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 4194304,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -389584,13 +389584,13 @@
     "hf_downloads": 46384,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 36,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 64,
+    "hidden_size": 2880,
+    "vocab_size": 201088,
+    "moe_intermediate_size": 2880,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -389910,7 +389910,7 @@
     "min_vram_gb": 60.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -389918,13 +389918,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 196608,
+    "moe_intermediate_size": 1280,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "upstage-solar-license"
@@ -389939,7 +389939,7 @@
     "min_vram_gb": 60.2,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -389947,14 +389947,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
+    "shared_expert_intermediate_size": 1024,
     "_discovered": true,
     "license": "openmdw-1.1"
   },
@@ -390294,7 +390294,7 @@
     "min_vram_gb": 60.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -390302,14 +390302,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
+    "shared_expert_intermediate_size": 1024,
     "_discovered": true,
     "license": "openmdw-1.1"
   },
@@ -390323,7 +390323,7 @@
     "min_vram_gb": 60.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -390331,14 +390331,14 @@
     "hf_downloads": 534,
     "hf_likes": 7,
     "release_date": "2026-07-21",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
+    "shared_expert_intermediate_size": 1024,
     "_discovered": true,
     "license": "openmdw-1.1"
   },
@@ -390352,7 +390352,7 @@
     "min_vram_gb": 60.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -390360,14 +390360,14 @@
     "hf_downloads": 722,
     "hf_likes": 2,
     "release_date": "2026-07-31",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 100352,
+    "moe_intermediate_size": 1024,
+    "shared_expert_intermediate_size": 1024,
     "_discovered": true,
     "license": "openmdw-1.1"
   },
@@ -392333,7 +392333,7 @@
     "min_vram_gb": 62.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -392341,14 +392341,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 3072,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 1024,
+    "shared_expert_intermediate_size": 1024,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -394190,7 +394190,7 @@
     "min_vram_gb": 65.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -394198,13 +394198,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 157184,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -394219,7 +394219,7 @@
     "min_vram_gb": 65.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -394227,13 +394227,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 42,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 32,
+    "head_dim": 128,
+    "hidden_size": 2560,
+    "vocab_size": 157184,
+    "moe_intermediate_size": 768,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -394863,7 +394863,7 @@
     "min_vram_gb": 76.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -394871,13 +394871,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "openrail",
@@ -394896,7 +394896,7 @@
     "min_vram_gb": 76.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -394904,13 +394904,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -394929,7 +394929,7 @@
     "min_vram_gb": 76.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "languages": [
@@ -394941,13 +394941,13 @@
     "hf_downloads": 155,
     "hf_likes": 2,
     "release_date": "2026-07-30",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 192,
+    "hidden_size": 6144,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -394962,7 +394962,7 @@
     "min_vram_gb": 78.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -394970,13 +394970,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -395127,7 +395127,7 @@
     "min_vram_gb": 81.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -395137,13 +395137,13 @@
     "hf_downloads": 312,
     "hf_likes": 10,
     "release_date": "2026-07-26",
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 192,
+    "hidden_size": 6144,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -395516,7 +395516,7 @@
     "min_vram_gb": 85.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -395524,13 +395524,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -395611,7 +395611,7 @@
     "min_vram_gb": 86.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -395619,13 +395619,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 192,
+    "hidden_size": 6144,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -395671,7 +395671,7 @@
     "min_vram_gb": 87.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -395679,13 +395679,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 120832,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "other"
@@ -395794,7 +395794,7 @@
     "min_vram_gb": 88.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -395802,13 +395802,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 120832,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -396052,7 +396052,7 @@
     "min_vram_gb": 90.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2048,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -396061,11 +396061,11 @@
     "hf_likes": 0,
     "release_date": null,
     "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
+    "num_attention_heads": 112,
+    "num_key_value_heads": 112,
     "head_dim": null,
     "hidden_size": null,
-    "vocab_size": null,
+    "vocab_size": 250880,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -396230,7 +396230,7 @@
     "min_vram_gb": 92.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -396238,13 +396238,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 53,
+    "num_attention_heads": 80,
+    "num_key_value_heads": 16,
+    "head_dim": 192,
+    "hidden_size": 4096,
+    "vocab_size": 220160,
+    "moe_intermediate_size": 1280,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -396414,7 +396414,7 @@
     "min_vram_gb": 101.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 512,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -396422,13 +396422,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 45,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 128815,
+    "moe_intermediate_size": 1280,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -397248,7 +397248,7 @@
     "min_vram_gb": 117.1,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 196608,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -397256,13 +397256,13 @@
     "hf_downloads": 2203,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 62,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 200064,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "minimax-m2.7-non-commercial",
@@ -398599,7 +398599,7 @@
     "min_vram_gb": 120.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 40960,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -398607,13 +398607,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 94,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -398632,7 +398632,7 @@
     "min_vram_gb": 120.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -398640,13 +398640,13 @@
     "hf_downloads": 1036,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 94,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 4,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0",
@@ -399939,7 +399939,7 @@
     "min_vram_gb": 123.5,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "Code generation and completion",
     "capabilities": [
       "tool_use"
@@ -399949,14 +399949,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 62,
+    "num_attention_heads": 96,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 151936,
+    "moe_intermediate_size": 2560,
+    "shared_expert_intermediate_size": 0,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -400141,7 +400141,7 @@
     "min_vram_gb": 129.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 2097152,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -400149,12 +400149,12 @@
     "hf_downloads": 2316,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
+    "num_hidden_layers": 162,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 128,
+    "head_dim": 128,
+    "hidden_size": 16384,
+    "vocab_size": 128256,
     "moe_intermediate_size": null,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
@@ -400401,7 +400401,7 @@
     "min_vram_gb": 139.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -400409,13 +400409,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 192,
+    "hidden_size": 6144,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -400462,7 +400462,7 @@
     "min_vram_gb": 145.6,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -400470,13 +400470,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -400726,7 +400726,7 @@
     "min_vram_gb": 145.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -400734,13 +400734,13 @@
     "hf_downloads": 3505,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -401681,7 +401681,7 @@
     "min_vram_gb": 151.1,
     "quantization": "Q4_K_M",
     "format": "mlx",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -401689,13 +401689,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 120832,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -401710,7 +401710,7 @@
     "min_vram_gb": 151.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -401718,13 +401718,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "vocab_size": 120832,
+    "moe_intermediate_size": 1536,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -402696,7 +402696,7 @@
     "min_vram_gb": 155.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -402704,13 +402704,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -403399,7 +403399,7 @@
     "min_vram_gb": 156.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -403407,13 +403407,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 43,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 4096,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -403604,7 +403604,7 @@
     "min_vram_gb": 159.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -403612,13 +403612,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 4,
+    "head_dim": 192,
+    "hidden_size": 4096,
+    "vocab_size": 152576,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -403814,7 +403814,7 @@
     "min_vram_gb": 163.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -403822,13 +403822,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 192,
+    "hidden_size": 6144,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -403874,7 +403874,7 @@
     "min_vram_gb": 171.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -403883,12 +403883,12 @@
     "hf_likes": 0,
     "release_date": null,
     "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 5120,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "openmdw-1.1"
@@ -403964,7 +403964,7 @@
     "min_vram_gb": 178.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -403974,13 +403974,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 192,
+    "hidden_size": 6144,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -404245,7 +404245,7 @@
     "min_vram_gb": 181.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 524288,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -404253,13 +404253,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 61,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 112,
+    "hidden_size": 7168,
+    "vocab_size": 163840,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -405358,7 +405358,7 @@
     "min_vram_gb": 203.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -405366,14 +405366,14 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
-    "shared_expert_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 2,
+    "head_dim": 256,
+    "hidden_size": 4096,
+    "vocab_size": 248320,
+    "moe_intermediate_size": 1024,
+    "shared_expert_intermediate_size": 1024,
     "_discovered": true,
     "license": "apache-2.0",
     "is_moe": true,
@@ -405617,7 +405617,7 @@
     "min_vram_gb": 204.2,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -405625,13 +405625,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 60,
+    "num_attention_heads": 48,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 3072,
+    "vocab_size": 200192,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "openmdw-1.1"
@@ -406215,7 +406215,7 @@
     "min_vram_gb": 207.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -406225,13 +406225,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 126,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 16384,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 53248,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -406407,7 +406407,7 @@
     "min_vram_gb": 207.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [
       "tool_use"
@@ -406417,13 +406417,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 126,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 16384,
+    "vocab_size": 128256,
+    "moe_intermediate_size": 53248,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "llama3"
@@ -406557,7 +406557,7 @@
     "min_vram_gb": 210.3,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -406565,13 +406565,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 153600,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -407239,7 +407239,7 @@
     "min_vram_gb": 233.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 10240000,
     "use_case": "Lightweight, edge deployment",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -407247,13 +407247,13 @@
     "hf_downloads": 13887,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 200064,
+    "moe_intermediate_size": 9216,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -408155,7 +408155,7 @@
     "min_vram_gb": 287.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -408164,12 +408164,12 @@
     "hf_likes": 0,
     "release_date": null,
     "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 5120,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "openmdw-1.1"
@@ -408184,7 +408184,7 @@
     "min_vram_gb": 287.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -408193,12 +408193,12 @@
     "hf_likes": 0,
     "release_date": null,
     "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 2,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 131072,
+    "moe_intermediate_size": 5120,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "openmdw-1.1"
@@ -408492,7 +408492,7 @@
     "min_vram_gb": 343.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 6553600,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -408500,13 +408500,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 61,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 128,
+    "head_dim": 64,
+    "hidden_size": 7168,
+    "vocab_size": 128815,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -408558,7 +408558,7 @@
     "min_vram_gb": 343.7,
     "quantization": "AWQ-4bit",
     "format": "awq",
-    "context_length": 4096,
+    "context_length": 6553600,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -408566,13 +408566,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 61,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 128,
+    "head_dim": 56,
+    "hidden_size": 7168,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -409211,7 +409211,7 @@
     "min_vram_gb": 350.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 6553600,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -409219,13 +409219,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 61,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 128,
+    "head_dim": 56,
+    "hidden_size": 7168,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "is_moe": true,
@@ -409444,7 +409444,7 @@
     "min_vram_gb": 351.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 6553600,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -409452,13 +409452,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 61,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 128,
+    "head_dim": 56,
+    "hidden_size": 7168,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -409473,7 +409473,7 @@
     "min_vram_gb": 351.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 6553600,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -409481,13 +409481,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 61,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 128,
+    "head_dim": 56,
+    "hidden_size": 7168,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -410184,7 +410184,7 @@
     "min_vram_gb": 383.8,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -410192,13 +410192,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 153600,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -410213,7 +410213,7 @@
     "min_vram_gb": 384.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 262144,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -410221,13 +410221,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 8,
+    "head_dim": 128,
+    "hidden_size": 6144,
+    "vocab_size": 153600,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "apache-2.0"
@@ -410407,7 +410407,7 @@
     "min_vram_gb": 385.9,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -410415,13 +410415,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 192,
+    "hidden_size": 6144,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -410981,7 +410981,7 @@
     "min_vram_gb": 386.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 202752,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -410989,13 +410989,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 78,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 64,
+    "hidden_size": 6144,
+    "vocab_size": 154880,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -411490,7 +411490,7 @@
     "min_vram_gb": 448.0,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "Multimodal, vision and text",
     "capabilities": [
       "vision"
@@ -411500,13 +411500,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 93,
+    "num_attention_heads": 96,
+    "num_key_value_heads": 96,
+    "head_dim": 74,
+    "hidden_size": 7168,
+    "vocab_size": 163840,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "kimi-k3"
@@ -411947,7 +411947,7 @@
     "min_vram_gb": 524.1,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -411955,13 +411955,13 @@
     "hf_downloads": 1804,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 70,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 8,
+    "head_dim": 192,
+    "hidden_size": 6144,
+    "vocab_size": 152576,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -412071,7 +412071,7 @@
     "min_vram_gb": 525.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 131072,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -412079,13 +412079,13 @@
     "hf_downloads": 1142,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 80,
+    "num_attention_heads": 64,
+    "num_key_value_heads": 64,
+    "head_dim": 128,
+    "hidden_size": 8192,
+    "vocab_size": 157184,
+    "moe_intermediate_size": 2048,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit"
@@ -413323,7 +413323,7 @@
     "min_vram_gb": 726.6,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 1048576,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -413331,13 +413331,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 93,
+    "num_attention_heads": 96,
+    "num_key_value_heads": 96,
+    "head_dim": 74,
+    "hidden_size": 7168,
+    "vocab_size": 163840,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true
   },
@@ -413415,7 +413415,7 @@
     "min_vram_gb": 805.7,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -413423,13 +413423,13 @@
     "hf_downloads": 7931,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 61,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 7168,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -413667,7 +413667,7 @@
     "min_vram_gb": 845.4,
     "quantization": "Q4_K_M",
     "format": "gguf",
-    "context_length": 4096,
+    "context_length": 16777216,
     "use_case": "General purpose text generation",
     "capabilities": [],
     "pipeline_tag": "text-generation",
@@ -413675,13 +413675,13 @@
     "hf_downloads": 0,
     "hf_likes": 0,
     "release_date": null,
-    "num_hidden_layers": null,
-    "num_attention_heads": null,
-    "num_key_value_heads": null,
-    "head_dim": null,
-    "hidden_size": null,
-    "vocab_size": null,
-    "moe_intermediate_size": null,
+    "num_hidden_layers": 61,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 1,
+    "head_dim": 512,
+    "hidden_size": 7168,
+    "vocab_size": 129280,
+    "moe_intermediate_size": 3072,
     "shared_expert_intermediate_size": null,
     "_discovered": true,
     "license": "mit",
@@ -414748,4 +414748,4 @@
     "_discovered": true,
     "license": "kimi-k3"
   }
-]
\ No newline at end of file
+]
diff --git a/llmfit-core/tests/schema_validation.rs b/llmfit-core/tests/schema_validation.rs
index f807e556..6f955e5b 100644
--- a/llmfit-core/tests/schema_validation.rs
+++ b/llmfit-core/tests/schema_validation.rs
@@ -42,3 +42,37 @@ fn hf_models_match_schema() {
 
     validate_file(&schema, "data/hf_models.json");
 }
+
+#[test]
+fn hf_models_keep_known_hybrid_attention_heads() {
+    // Regression for the 2026-08-28 weekly scrape: a missed config.json fetch
+    // nullified architecture fields, and "mamba" in the name then erased KV.
+    // Weekly CI runs `hf_models_*`; keep this name so it gates the refresh PR.
+    let path = Path::new(env!("CARGO_MANIFEST_DIR")).join("data/hf_models.json");
+    let text = std::fs::read_to_string(&path)
+        .unwrap_or_else(|e| panic!("cannot read {}: {e}", path.display()));
+    let data: Value = serde_json::from_str(&text)
+        .unwrap_or_else(|e| panic!("cannot parse {}: {e}", path.display()));
+    let models = data.as_array().expect("hf_models.json is an array");
+    let required = [(
+        "QwerkyAI/Qwerky-Optimized-Llama3.2-Mamba-0.2-3B-Instruct",
+        24u64,
+        8u64,
+    )];
+    for (name, attn, kv) in required {
+        let model = models
+            .iter()
+            .find(|m| m.get("name").and_then(|v| v.as_str()) == Some(name))
+            .unwrap_or_else(|| panic!("missing {name} in embedded catalog"));
+        assert_eq!(
+            model.get("num_attention_heads").and_then(|v| v.as_u64()),
+            Some(attn),
+            "{name} lost num_attention_heads"
+        );
+        assert_eq!(
+            model.get("num_key_value_heads").and_then(|v| v.as_u64()),
+            Some(kv),
+            "{name} lost num_key_value_heads"
+        );
+    }
+}
diff --git a/scripts/scrape_hf_models.py b/scripts/scrape_hf_models.py
index 1aafa4e7..29a9d610 100644
--- a/scripts/scrape_hf_models.py
+++ b/scripts/scrape_hf_models.py
@@ -467,6 +467,61 @@ def estimate_vram(total_params: int, quant: str) -> float:
     return round(max(vram_gb, 0.5), 1)
 
 
+
+# Architecture fields filled from config.json. A failed fetch returns all-null
+# and used to overwrite good catalog values on the weekly merge. Keep the old
+# numbers when the fresh scrape did not produce them.
+ARCH_METADATA_KEYS = (
+    "num_hidden_layers",
+    "num_attention_heads",
+    "num_key_value_heads",
+    "head_dim",
+    "hidden_size",
+    "vocab_size",
+    "moe_intermediate_size",
+    "shared_expert_intermediate_size",
+)
+
+# A missed config.json fetch falls back to 4096. Restoring prior context is
+# only safe when architecture metadata was also dropped in the same pass.
+DEFAULT_CONTEXT_LENGTH = 4096
+
+# Mass config-fetch failure. One or two models can lose config.json for real;
+# 25+ in a single scrape is the 2026-08-28 incident (1,764 models).
+ARCH_METADATA_DROP_LIMIT = 25
+
+
+def preserve_existing_metadata(old_model: dict, fresh_model: dict) -> list[str]:
+    """Copy catalog fields the fresh scrape dropped. Returns restored keys.
+
+    Mutates ``fresh_model``. Used both to keep the weekly merge additive and
+    to count how many models needed an architecture-metadata rescue.
+    """
+    restored: list[str] = []
+    if old_model.get("license") and not fresh_model.get("license"):
+        fresh_model["license"] = old_model["license"]
+        restored.append("license")
+    if old_model.get("gguf_sources") and not fresh_model.get("gguf_sources"):
+        fresh_model["gguf_sources"] = old_model["gguf_sources"]
+        restored.append("gguf_sources")
+    for key in ("hf_downloads", "hf_likes", "release_date", "languages"):
+        if old_model.get(key) and not fresh_model.get(key):
+            fresh_model[key] = old_model[key]
+            restored.append(key)
+    for key in ARCH_METADATA_KEYS:
+        if old_model.get(key) is not None and fresh_model.get(key) is None:
+            fresh_model[key] = old_model[key]
+            restored.append(key)
+    if (
+        fresh_model.get("context_length") == DEFAULT_CONTEXT_LENGTH
+        and old_model.get("context_length") not in (None, DEFAULT_CONTEXT_LENGTH)
+        and "num_attention_heads" in restored
+    ):
+        fresh_model["context_length"] = old_model["context_length"]
+        restored.append("context_length")
+    return restored
+
+
 def extract_arch_metadata(config: dict | None) -> dict:
     """Extract architecture fields for precise KV cache and MoE speed estimation.
 
@@ -2972,6 +3027,7 @@ def main():
     existing_count = 0
     retained_count = 0
     updated_count = 0
+    arch_heads_restored = 0
     for output_path in output_paths:
         if os.path.exists(output_path):
             try:
@@ -2982,16 +3038,9 @@ def main():
                     name = old_model.get("name", "")
                     if name in fresh_by_name:
                         fresh_model = fresh_by_name[name]
-                        if old_model.get("license") and not fresh_model.get("license"):
-                            fresh_model["license"] = old_model["license"]
-                        if old_model.get("gguf_sources") and not fresh_model.get("gguf_sources"):
-                            fresh_model["gguf_sources"] = old_model["gguf_sources"]
-                        # Fallback stubs and trending listings carry no
-                        # popularity/date/language metadata — never let them
-                        # clobber real values from a previous scrape.
-                        for key in ("hf_downloads", "hf_likes", "release_date", "languages"):
-                            if old_model.get(key) and not fresh_model.get(key):
-                                fresh_model[key] = old_model[key]
+                        restored = preserve_existing_metadata(old_model, fresh_model)
+                        if "num_attention_heads" in restored:
+                            arch_heads_restored += 1
                         updated_count += 1
                     elif name:
                         # Historical model not in current scrape — keep it
@@ -3006,6 +3055,15 @@ def main():
     if existing_count:
         print(f"\nMerged with existing database ({existing_count} models):")
         print(f"  Updated: {updated_count}, Retained historical: {retained_count}")
+        print(f"  Architecture heads restored from prior catalog: {arch_heads_restored}")
+        if arch_heads_restored > ARCH_METADATA_DROP_LIMIT:
+            print(
+                f"ERROR: config.json fetch dropped attention-head metadata for "
+                f"{arch_heads_restored} models (limit {ARCH_METADATA_DROP_LIMIT}). "
+                f"Prior values were kept, but refusing to ship this scrape so the "
+                f"mass drop cannot land quietly again."
+            )
+            sys.exit(1)
 
     # Keep additive/retained entries on the current schema even if they were
     # produced by an older scraper version.
diff --git a/scripts/test_preserve_catalog_metadata.py b/scripts/test_preserve_catalog_metadata.py
new file mode 100644
index 00000000..b487cb32
--- /dev/null
+++ b/scripts/test_preserve_catalog_metadata.py
@@ -0,0 +1,61 @@
+#!/usr/bin/env python3
+"""Guard: weekly scrape must not silently wipe architecture metadata."""
+
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent))
+from scrape_hf_models import (  # noqa: E402
+    ARCH_METADATA_DROP_LIMIT,
+    preserve_existing_metadata,
+)
+
+
+def test_preserves_architecture_when_config_fetch_misses():
+    old = {
+        "license": "mit",
+        "num_attention_heads": 24,
+        "num_key_value_heads": 8,
+        "num_hidden_layers": 28,
+        "context_length": 4194304,
+        "hf_downloads": 89,
+    }
+    fresh = {
+        "license": None,
+        "num_attention_heads": None,
+        "num_key_value_heads": None,
+        "num_hidden_layers": None,
+        "context_length": 4096,
+        "hf_downloads": 0,
+    }
+    restored = preserve_existing_metadata(old, fresh)
+    assert fresh["num_attention_heads"] == 24, restored
+    assert fresh["num_key_value_heads"] == 8, restored
+    assert fresh["num_hidden_layers"] == 28, restored
+    assert fresh["context_length"] == 4194304, restored
+    assert fresh["license"] == "mit", restored
+    assert "num_attention_heads" in restored
+
+
+def test_does_not_invent_heads_the_catalog_never_had():
+    old = {"num_attention_heads": None, "context_length": 2048}
+    fresh = {"num_attention_heads": None, "context_length": 4096}
+    preserve_existing_metadata(old, fresh)
+    assert fresh["num_attention_heads"] is None
+    assert fresh["context_length"] == 4096
+
+
+def test_mass_drop_limit_would_have_caught_2026_08_28():
+    assert ARCH_METADATA_DROP_LIMIT < 1764
+
+
+if __name__ == "__main__":
+    tests = [
+        test_preserves_architecture_when_config_fetch_misses,
+        test_does_not_invent_heads_the_catalog_never_had,
+        test_mass_drop_limit_would_have_caught_2026_08_28,
+    ]
+    for fn in tests:
+        fn()
+        print(f"ok  {fn.__name__}")
+    print(f"{len(tests)} passed")