@bendyline/gilde 0.1.14 → 0.1.16

package/data/chat-models/bt/btl4-35b-q4/manifest.json

@@ -0,0 +1,147 @@
+{
+  "schemaVersion": 1,
+  "kind": "chat-model",
+  "id": "btl4-35b-q4",
+  "name": "BTL-4 (35B-A3B, Q4)",
+  "description": "Bad Theory Labs' agentic fine-tune of Ornith 1.0 35B, at 35B total parameters with ~2B active per token (Mixture-of-Experts: 256 experts, 8 routed). 30 of its 40 layers use linear attention, so the KV cache costs only ~20 KB/token and the full 256K context stays affordable. Q4_K_M quantization from bartowski — the full-quality companion to the 2.3-bit Compact build. The publisher's benchmark claims are not independently verified.",
+  "tags": [
+    "badtheorylabs",
+    "agentic",
+    "tools",
+    "code",
+    "reasoning",
+    "long-context",
+    "moe"
+  ],
+  "category": "coding",
+  "maintainer": {
+    "name": "Bad Theory Labs",
+    "url": "https://huggingface.co/badtheorylabs"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-08-06T00:00:00Z",
+  "license": "Apache-2.0",
+  "licenseClass": "open",
+  "licenseShortName": "Apache 2.0",
+  "recoScore": 10,
+  "parameterSize": "35B",
+  "approxSizeBytes": 21391448192,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "kvBytesPerTokenF16": 20480,
+  "kvFixedBytesF16": 64389120,
+  "upstream": "https://huggingface.co/badtheorylabs/BTL-4",
+  "style": {
+    "family": "qwen",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.private-reasoning-guidance",
+    "prompt.tool-cookbook-condensed",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.ollama-num-predict-bumped",
+    "turn.preamble-folding",
+    "mcp.compact-tool-schemas"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 0.6,
+      "topP": 0.95,
+      "topK": 20,
+      "minP": 0,
+      "maxTokens": 16384
+    },
+    "samplingWhenThinking": {
+      "temperature": 0.6,
+      "topP": 0.95
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 8192
+    },
+    "promptTags": {
+      "enableThinkingTag": "/think",
+      "disableThinkingTag": "/no_think"
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 24576
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 8192
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 32768
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 12288
+        }
+      },
+      "thinking-precise": {
+        "sampling": {
+          "temperature": 0.4,
+          "topP": 0.9,
+          "topK": 20,
+          "minP": 0,
+          "seed": 0,
+          "maxTokens": 6144
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "instruct": {
+        "sampling": {
+          "temperature": 0.7,
+          "topP": 0.8,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 1.5,
+          "repetitionPenalty": 1,
+          "maxTokens": 4096
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      },
+      "creative": {
+        "sampling": {
+          "temperature": 1.1,
+          "topP": 0.95,
+          "topK": 40,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      }
+    }
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "bartowski/badtheorylabs_BTL-4-GGUF",
+    "revision": "799d011b498e3287f514eb463ea04d6624ec945f",
+    "filename": "badtheorylabs_BTL-4-Q4_K_M.gguf",
+    "sha256": "75ce22e4f603bdf3f94c2b32151bb6d3492d2ab8a34ad8a4bfa2eaef0125c50f",
+    "approxSizeBytes": 21391448192,
+    "quantization": "Q4_K_M",
+    "residentBytes": 25669737830
+  }
+}

package/data/chat-models/bt/btl4-35b-q4/versions/1.0.0/manifest.json

@@ -0,0 +1,15 @@
+{
+  "schemaVersion": 1,
+  "version": "1.0.0",
+  "releasedAt": "2026-08-07T00:00:00Z",
+  "approxSizeBytes": 21391448192,
+  "llamaCpp": {
+    "huggingfaceRepo": "bartowski/badtheorylabs_BTL-4-GGUF",
+    "revision": "799d011b498e3287f514eb463ea04d6624ec945f",
+    "filename": "badtheorylabs_BTL-4-Q4_K_M.gguf",
+    "sha256": "75ce22e4f603bdf3f94c2b32151bb6d3492d2ab8a34ad8a4bfa2eaef0125c50f",
+    "approxSizeBytes": 21391448192,
+    "quantization": "Q4_K_M",
+    "residentBytes": 25669737830
+  }
+}

package/data/chat-models/bt/btl4-compact-iq2/manifest.json

@@ -0,0 +1,147 @@
+{
+  "schemaVersion": 1,
+  "kind": "chat-model",
+  "id": "btl4-compact-iq2",
+  "name": "BTL-4 Compact (35B-A3B, IQ2_XXS)",
+  "description": "Bad Theory Labs' agentic fine-tune of Ornith 1.0 35B, at 35B total parameters with ~2B active per token (Mixture-of-Experts: 256 experts, 8 routed). 30 of its 40 layers use linear attention, so the KV cache costs only ~20 KB/token and the full 256K context stays affordable on a laptop. This is a 2.3 bit-per-weight imatrix quantization: expert tensors at IQ2_XXS, everything else at Q4_K_M, which fits the whole model in under 10 GB. The publisher's benchmark claims are not independently verified.",
+  "tags": [
+    "badtheorylabs",
+    "agentic",
+    "tools",
+    "code",
+    "reasoning",
+    "long-context",
+    "moe"
+  ],
+  "category": "coding",
+  "maintainer": {
+    "name": "Bad Theory Labs",
+    "url": "https://huggingface.co/badtheorylabs"
+  },
+  "version": "1.0.1",
+  "updatedAt": "2026-08-06T00:00:00Z",
+  "license": "Apache-2.0",
+  "licenseClass": "open",
+  "licenseShortName": "Apache 2.0",
+  "recoScore": 10,
+  "parameterSize": "35B",
+  "approxSizeBytes": 9967966240,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "kvBytesPerTokenF16": 20480,
+  "kvFixedBytesF16": 64389120,
+  "upstream": "https://huggingface.co/badtheorylabs/BTL-4-Compact",
+  "style": {
+    "family": "qwen",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.private-reasoning-guidance",
+    "prompt.tool-cookbook-condensed",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.ollama-num-predict-bumped",
+    "turn.preamble-folding",
+    "mcp.compact-tool-schemas"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 0.6,
+      "topP": 0.95,
+      "topK": 20,
+      "minP": 0,
+      "maxTokens": 16384
+    },
+    "samplingWhenThinking": {
+      "temperature": 0.6,
+      "topP": 0.95
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 8192
+    },
+    "promptTags": {
+      "enableThinkingTag": "/think",
+      "disableThinkingTag": "/no_think"
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 24576
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 8192
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 32768
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 12288
+        }
+      },
+      "thinking-precise": {
+        "sampling": {
+          "temperature": 0.4,
+          "topP": 0.9,
+          "topK": 20,
+          "minP": 0,
+          "seed": 0,
+          "maxTokens": 6144
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "instruct": {
+        "sampling": {
+          "temperature": 0.7,
+          "topP": 0.8,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 1.5,
+          "repetitionPenalty": 1,
+          "maxTokens": 4096
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      },
+      "creative": {
+        "sampling": {
+          "temperature": 1.1,
+          "topP": 0.95,
+          "topK": 40,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      }
+    }
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "badtheorylabs/BTL-4-Compact",
+    "revision": "2a29bc80b0ebc80d6409fbad89f85d910dbb4aee",
+    "filename": "BTL-4-IQ2_XXS.gguf",
+    "sha256": "6b7c298cf909fc04428ecf360a29dcc578188b1c90aa6ed435159f5a0d351496",
+    "approxSizeBytes": 9967966240,
+    "quantization": "IQ2_XXS",
+    "residentBytes": 11961559488
+  }
+}

package/data/chat-models/bt/btl4-compact-iq2/versions/1.0.0/manifest.json

@@ -0,0 +1,15 @@
+{
+  "schemaVersion": 1,
+  "version": "1.0.0",
+  "releasedAt": "2026-08-06T00:00:00Z",
+  "approxSizeBytes": 9967966240,
+  "llamaCpp": {
+    "huggingfaceRepo": "badtheorylabs/BTL-4-Compact",
+    "revision": "2a29bc80b0ebc80d6409fbad89f85d910dbb4aee",
+    "filename": "BTL-4-IQ2_XXS.gguf",
+    "sha256": "6b7c298cf909fc04428ecf360a29dcc578188b1c90aa6ed435159f5a0d351496",
+    "approxSizeBytes": 9967966240,
+    "quantization": "IQ2_XXS",
+    "residentBytes": 11961559488
+  }
+}

package/data/chat-models/bt/btl4-compact-iq2/versions/1.0.1/manifest.json

@@ -0,0 +1,15 @@
+{
+  "schemaVersion": 1,
+  "version": "1.0.1",
+  "releasedAt": "2026-08-06T00:00:00Z",
+  "approxSizeBytes": 9967966240,
+  "llamaCpp": {
+    "huggingfaceRepo": "badtheorylabs/BTL-4-Compact",
+    "revision": "2a29bc80b0ebc80d6409fbad89f85d910dbb4aee",
+    "filename": "BTL-4-IQ2_XXS.gguf",
+    "sha256": "6b7c298cf909fc04428ecf360a29dcc578188b1c90aa6ed435159f5a0d351496",
+    "approxSizeBytes": 9967966240,
+    "quantization": "IQ2_XXS",
+    "residentBytes": 11961559488
+  }
+}

package/data/chat-models/ge/gemma4-12b-q4/manifest.json

@@ -81,11 +81,6 @@
       "topK": 64,
       "repetitionPenalty": 1.1,
       "repetitionContext": 64,
-      "dry": {
-        "multiplier": 0.8,
-        "base": 1.75,
-        "allowedLength": 2
-      },
       "maxTokens": 8192
     },
     "reasoning": {
@@ -100,11 +95,6 @@
           "topK": 64,
           "repetitionPenalty": 1.1,
           "repetitionContext": 64,
-          "dry": {
-            "multiplier": 0.8,
-            "base": 1.75,
-            "allowedLength": 2
-          },
           "maxTokens": 2048
         }
       },
@@ -115,11 +105,6 @@
           "topK": 40,
           "repetitionPenalty": 1.05,
           "repetitionContext": 64,
-          "dry": {
-            "multiplier": 0.8,
-            "base": 1.75,
-            "allowedLength": 2
-          },
           "maxTokens": 2048
         }
       },
@@ -130,11 +115,6 @@
           "topK": 20,
           "repetitionPenalty": 1.05,
           "repetitionContext": 64,
-          "dry": {
-            "multiplier": 0.8,
-            "base": 1.75,
-            "allowedLength": 2
-          },
           "maxTokens": 4096,
           "seed": 0
         }
@@ -146,11 +126,6 @@
           "topK": 64,
           "repetitionPenalty": 1.1,
           "repetitionContext": 64,
-          "dry": {
-            "multiplier": 0.8,
-            "base": 1.75,
-            "allowedLength": 2
-          },
           "maxTokens": 4096
         }
       },