@bendyline/gilde 0.1.18 → 0.1.20

package/authoring/chat-models/ternary-bonsai-27b-q2.json

@@ -0,0 +1,144 @@
+{
+  "id": "ternary-bonsai-27b-q2",
+  "name": "Ternary Bonsai 27B (2-bit)",
+  "description": "PrismML's ternary 2-bit adaptation of Qwen 3.6 27B. Multimodal (text + image + video), 262K native context, thinking and tool-calling support, packaged for Apple Silicon in roughly 8.5GB. A promising 27B-class on-device option for Macs with 16GB+ unified memory; 24GB+ is recommended for useful context headroom.",
+  "tags": [
+    "prismml",
+    "qwen",
+    "ternary",
+    "2-bit",
+    "multimodal",
+    "vision",
+    "video",
+    "tools",
+    "reasoning",
+    "long-context"
+  ],
+  "category": "general",
+  "maintainer": {
+    "name": "PrismML",
+    "url": "https://prismml.com"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-07-14T00:00:00Z",
+  "license": "Apache-2.0",
+  "licenseClass": "open",
+  "licenseShortName": "Apache 2.0",
+  "parameterSize": "27B",
+  "approxSizeBytes": 8520000000,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "upstream": "https://huggingface.co/prism-ml/Ternary-Bonsai-27B-mlx-2bit",
+  "style": {
+    "family": "qwen",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "provider.merge-system-messages",
+    "mcp.compact-tool-schemas",
+    "provider.compact-write-transcript",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.ollama-num-predict-bumped",
+    "turn.preamble-folding",
+    "turn.ramble-detection",
+    "tools.mlx-grammar",
+    "prompt.derive-by-execution"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 0.7,
+      "topP": 0.95,
+      "topK": 20,
+      "minP": 0,
+      "maxTokens": 8192
+    },
+    "samplingWhenThinking": {
+      "temperature": 0.7,
+      "topP": 0.95
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 4096
+    },
+    "promptTags": {
+      "enableThinkingTag": "/think",
+      "disableThinkingTag": "/no_think"
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 0.7,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 12288
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 6144
+        }
+      },
+      "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
+        }
+      }
+    }
+  },
+  "mlx": {
+    "huggingfaceRepo": "prism-ml/Ternary-Bonsai-27B-mlx-2bit",
+    "quantization": "2bit",
+    "residentBytes": 11100000000
+  }
+}

package/data/chat-models/gl/glm-5.2-754b-q2/manifest.json

@@ -0,0 +1,65 @@
+{
+  "schemaVersion": 1,
+  "kind": "chat-model",
+  "id": "glm-5.2-754b-q2",
+  "name": "GLM 5.2 (IQ2_XXS)",
+  "description": "antirez's routed 2-bit community GGUF of Z.ai's GLM 5.2 — a 754B-parameter Mixture-of-Experts (256 experts, 8 active per token, ~40B active) with a 1M-token context and DeepSeek-style sparse attention. The routed experts are crushed to IQ2_XXS (block 78 in Q2_K) while attention, the shared expert, embeddings, and the output head stay at Q8-class precision, landing the file at ~197 GB. Only ~20 GB of that is non-routed weight that must stay resident: DwarfStar streams the remaining ~177 GB of experts from SSD, so it runs on a 128 GB machine — but it is a very large download and a slow first turn. Built for antirez's `ds4` engine; these quants do not load on stock llama.cpp.",
+  "tags": [
+    "glm",
+    "reasoning",
+    "moe",
+    "tools",
+    "2-bit",
+    "iq2",
+    "long-context",
+    "large"
+  ],
+  "category": "reasoning",
+  "maintainer": {
+    "name": "antirez",
+    "url": "https://huggingface.co/antirez/glm-5.2-gguf"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-07-25T00:00:00Z",
+  "license": "MIT",
+  "parameterSize": "754B",
+  "approxSizeBytes": 211075856448,
+  "supportsTools": true,
+  "contextWindow": 1048576,
+  "upstream": "https://huggingface.co/zai-org/GLM-5.2",
+  "style": {
+    "family": "glm",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.tool-cookbook-condensed",
+    "mcp.compact-tool-schemas",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.preamble-folding"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 1,
+      "topP": 0.95,
+      "maxTokens": 16384
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 4096
+    }
+  },
+  "ds4": {
+    "huggingfaceRepo": "antirez/glm-5.2-gguf",
+    "revision": "2638b3b878f5c6cc3ae7334b8dbea1275025f52e",
+    "filename": "GLM-5.2-UD-IQ2_XXS_RoutedIQ2XXS_blk78Q2K.gguf",
+    "sha256": "a49de64c5020432bdae23de36a423a9660a5621bc0db8d12b66bd8814b07fea0",
+    "approxSizeBytes": 211075856448,
+    "quantization": "IQ2_XXS",
+    "residentBytes": 61363217408,
+    "cacheExpertsBytes": 34359738368,
+    "ssdStreaming": true,
+    "maxLaunchCtx": 65536
+  }
+}

package/data/chat-models/gl/glm-5.2-754b-q2/versions/1.0.0/manifest.json

@@ -0,0 +1,18 @@
+{
+  "schemaVersion": 1,
+  "version": "1.0.0",
+  "releasedAt": "2026-07-25T00:00:00Z",
+  "approxSizeBytes": 211075856448,
+  "ds4": {
+    "huggingfaceRepo": "antirez/glm-5.2-gguf",
+    "revision": "2638b3b878f5c6cc3ae7334b8dbea1275025f52e",
+    "filename": "GLM-5.2-UD-IQ2_XXS_RoutedIQ2XXS_blk78Q2K.gguf",
+    "sha256": "a49de64c5020432bdae23de36a423a9660a5621bc0db8d12b66bd8814b07fea0",
+    "approxSizeBytes": 211075856448,
+    "quantization": "IQ2_XXS",
+    "residentBytes": 61363217408,
+    "cacheExpertsBytes": 34359738368,
+    "ssdStreaming": true,
+    "maxLaunchCtx": 65536
+  }
+}

package/data/chat-models/gp/gpt-oss-120b-q4/manifest.json

@@ -38,6 +38,11 @@
     "fabrication.detect-past-tense-no-tools",
     "turn.preamble-folding"
   ],
+  "tuning": {
+    "reasoning": {
+      "effort": "medium"
+    }
+  },
   "ollama": {
     "tag": "gpt-oss:120b"
   },
@@ -166,10 +171,5 @@
         "sizeBytes": 4200
       }
     ]
-  },
-  "tuning": {
-    "reasoning": {
-      "effort": "medium"
-    }
   }
 }

package/data/chat-models/gp/gpt-oss-120b-q4/versions/1.1.0/manifest.json

@@ -0,0 +1,135 @@
+{
+  "schemaVersion": 1,
+  "version": "1.1.0",
+  "releasedAt": "2026-05-17T00:00:00Z",
+  "approxSizeBytes": 65000000000,
+  "ollama": {
+    "tag": "gpt-oss:120b"
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "bartowski/openai_gpt-oss-120b-GGUF",
+    "revision": "8d64014efc314685c68ad448590f905fcb20e534",
+    "shards": [
+      {
+        "name": "openai_gpt-oss-120b-Q4_K_M/openai_gpt-oss-120b-Q4_K_M-00001-of-00002.gguf",
+        "sha256": "48b6bea561b52aaa3c10ca7681f35a923196cc348098e55ffc6ee3d184eb0949",
+        "sizeBytes": 39954683808
+      },
+      {
+        "name": "openai_gpt-oss-120b-Q4_K_M/openai_gpt-oss-120b-Q4_K_M-00002-of-00002.gguf",
+        "sha256": "06ab40ac3a9e2d7a77b6491b4cf76af152bee88ce493331a1f5147a0207abcad",
+        "sizeBytes": 22887029536
+      }
+    ],
+    "approxSizeBytes": 62841713344,
+    "quantization": "Q4_K_M"
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/gpt-oss-120b-MXFP4-Q4",
+    "revision": "bce781bef0f2fc85ed4e575af74054f5aad73ddd",
+    "quantization": "MXFP4-Q4",
+    "approxSizeBytes": 62358097894,
+    "files": [
+      {
+        "name": "chat_template.jinja",
+        "sha256": "a4c9919cbbd4acdd51ccffe22da049264b1b73e59055fa58811a99efbd7c8146",
+        "sizeBytes": 16738
+      },
+      {
+        "name": "config.json",
+        "sha256": "d15a4db435befc6d28eff75c34f25386fb77ae56b88630857f372fd1fc25289d",
+        "sizeBytes": 50145
+      },
+      {
+        "name": "generation_config.json",
+        "sha256": "f9970ada892d2d1f72e3ed0a6535ccebadd11897318794ca671d8c7014c957da",
+        "sizeBytes": 177
+      },
+      {
+        "name": "model-00001-of-00013.safetensors",
+        "sha256": "f253063b8f2856e1e4bc0b3c3dcae743308574786c3de8e2b34e6108439a92a1",
+        "sizeBytes": 4930648083
+      },
+      {
+        "name": "model-00002-of-00013.safetensors",
+        "sha256": "75e5e879cf3eb84d8d520dd084f60aa9fa804a8ed9fe5bbfa3646c0abc196fcd",
+        "sizeBytes": 5133844451
+      },
+      {
+        "name": "model-00003-of-00013.safetensors",
+        "sha256": "8d45b4bd9a13913f1422d8056ab5e5455d961ee342810a483c65bc50e13e9825",
+        "sizeBytes": 5133844463
+      },
+      {
+        "name": "model-00004-of-00013.safetensors",
+        "sha256": "58a266ddeb125b5bb2f62329c1db6af5907917bce7e3228b6cd4c836d2b6ff1d",
+        "sizeBytes": 5133844452
+      },
+      {
+        "name": "model-00005-of-00013.safetensors",
+        "sha256": "9660b5a6a42ec1ecf59dd2d437ebb8f1e145d27d3d1a3b6551e4ad5f1231251c",
+        "sizeBytes": 5133844583
+      },
+      {
+        "name": "model-00006-of-00013.safetensors",
+        "sha256": "6a25c6255571f094de71bfbbbe0788edf40a1f42732caa67d5e34cafabe2b483",
+        "sizeBytes": 5133844503
+      },
+      {
+        "name": "model-00007-of-00013.safetensors",
+        "sha256": "baf2929ba19f5237b2eee1d70c62d366b7fe7efd2a36c301b276eb05117f92f4",
+        "sizeBytes": 5133844579
+      },
+      {
+        "name": "model-00008-of-00013.safetensors",
+        "sha256": "851ffd789b3fb575c93441b99a41326a1901ca931bebd1547fce68df91f6f9ae",
+        "sizeBytes": 5133844553
+      },
+      {
+        "name": "model-00009-of-00013.safetensors",
+        "sha256": "491729509a6b125c68f6756bb14fc27be8dc264fae3090ed6422468a07c16080",
+        "sizeBytes": 5133844551
+      },
+      {
+        "name": "model-00010-of-00013.safetensors",
+        "sha256": "3a985b1bb4073449483edcd1058390d46d6c2be2db74f77681ecd86114190367",
+        "sizeBytes": 5133844567
+      },
+      {
+        "name": "model-00011-of-00013.safetensors",
+        "sha256": "455f0edf416082de29bbcd906869bfcb2b8fddab58c3d6e57e348c72ccda0056",
+        "sizeBytes": 5133844583
+      },
+      {
+        "name": "model-00012-of-00013.safetensors",
+        "sha256": "40c8cf3a1797dce2b90a7e1db5858e337c173fc0e9055efe9362c56b8e0f06f1",
+        "sizeBytes": 5133844567
+      },
+      {
+        "name": "model-00013-of-00013.safetensors",
+        "sha256": "93e1477c64c3bd08f4f4202aee1e7305187e86dc82666659435620317f89bd83",
+        "sizeBytes": 927119654
+      },
+      {
+        "name": "model.safetensors.index.json",
+        "sha256": "6b2aab23124d1b9761c64a4f0076c8ecaf8955ed1d09f217aa9fb82c568991a6",
+        "sizeBytes": 100431
+      },
+      {
+        "name": "special_tokens_map.json",
+        "sha256": "8464cabd6eda239fe46ebf8ae63b46c417721784a961a022f6b59174a2cda0e2",
+        "sizeBytes": 440
+      },
+      {
+        "name": "tokenizer.json",
+        "sha256": "0614fe83cadab421296e664e1f48f4261fa8fef6e03e63bb75c20f38e37d07d3",
+        "sizeBytes": 27868174
+      },
+      {
+        "name": "tokenizer_config.json",
+        "sha256": "9279e942392b742d633c7adbb89ebe002c98399db8926a7af5125c726f404070",
+        "sizeBytes": 4200
+      }
+    ]
+  }
+}