@bendyline/gilde 0.1.55 → 0.1.57

package/authoring/chat-models/glm-5.3-flash-320b-q2.json

@@ -0,0 +1,129 @@
+{
+  "id": "glm-5.3-flash-320b-q2",
+  "name": "GLM 5.3 Flash (320B-A18B, IQ2_XXS)",
+  "description": "antirez's routed 2-bit GGUF of Z.ai's GLM 5.3 Flash — a natively multimodal 320B-parameter Mixture-of-Experts with 18B active parameters, a 1M-token context, recurrent KDA layers, sparse DSA attention, hyper-connections, and a built-in MTP draft block. The Q2 file is about 90 GiB and targets a 128 GB Mac or DGX Spark with a modest initial context. Its separately packaged vision encoder adds image understanding to the same text model when Gezel launches ds4 with `--vision`.",
+  "tags": [
+    "glm",
+    "zai",
+    "agentic",
+    "coding",
+    "multimodal",
+    "vision",
+    "reasoning",
+    "moe",
+    "tools",
+    "2-bit",
+    "iq2",
+    "long-context",
+    "large"
+  ],
+  "category": "reasoning",
+  "maintainer": {
+    "name": "antirez",
+    "url": "https://huggingface.co/antirez/glm-5.3-flash-gguf"
+  },
+  "maker": {
+    "name": "Z.ai",
+    "url": "https://huggingface.co/zai-org/GLM-5.3-Flash"
+  },
+  "version": "1.0.0",
+  "versionMinGezelVersion": "1.26248",
+  "updatedAt": "2026-09-06T00:00:00Z",
+  "license": "MIT",
+  "licenseClass": "open",
+  "licenseShortName": "MIT",
+  "parameterSize": "320B",
+  "approxSizeBytes": 97633097344,
+  "supportsTools": true,
+  "contextWindow": 1048576,
+  "upstream": "https://huggingface.co/zai-org/GLM-5.3-Flash",
+  "style": {
+    "family": "glm",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.tool-cookbook-condensed",
+    "prompt.meester-build-prelude",
+    "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
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "thinking-precise": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "seed": 0,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "instruct": {
+        "sampling": {
+          "temperature": 0.7,
+          "topP": 1,
+          "maxTokens": 4096
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      },
+      "creative": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      }
+    }
+  },
+  "ds4": {
+    "huggingfaceRepo": "antirez/glm-5.3-flash-gguf",
+    "filename": "GLM-5.3-Flash-Q2.gguf",
+    "visionEncoderFilename": "GLM-5.3-Flash-Vision-Encoder.gguf",
+    "quantization": "IQ2_XXS/Q2_K",
+    "residentBytes": 100865554432,
+    "ssdStreaming": false,
+    "maxLaunchCtx": 32768
+  }
+}

package/authoring/chat-models/qwen3.8-flash-next-iq3.json

@@ -0,0 +1,198 @@
+{
+  "id": "qwen3.8-flash-next-iq3",
+  "name": "Qwen 3.8 Flash Next (180B-A6B, IQ3_XXS)",
+  "description": "Qwen's experimental preview of the architecture behind Qwen4: a 125B-parameter MoE with 6B active per token, plus a 51B n-gram embedding table designed for cheap offload. This Unsloth UD-IQ3_XXS build is an 82 GB split GGUF whose 26.8 GiB n-gram table can be demand-paged; it is the practical variant for 64 GB system RAM plus a 32 GB-class discrete GPU. Gezel starts it conservatively at 64K context and one slot, then chooses the expert GPU/CPU split from the machine's measured VRAM.",
+  "tags": [
+    "qwen",
+    "agentic",
+    "coding",
+    "multimodal",
+    "vision",
+    "video",
+    "tools",
+    "reasoning",
+    "long-context",
+    "moe",
+    "ngram-embedding",
+    "large",
+    "experimental"
+  ],
+  "category": "general",
+  "maintainer": {
+    "name": "Unsloth",
+    "url": "https://huggingface.co/unsloth/Qwen3.8-Flash-Next-GGUF"
+  },
+  "maker": {
+    "name": "Qwen",
+    "url": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next"
+  },
+  "version": "1.0.0",
+  "versionMinGezelVersion": "1.26248",
+  "updatedAt": "2026-09-06T00:00:00Z",
+  "license": "qwen-community-1.0",
+  "licenseClass": "custom-restricted",
+  "licenseShortName": "Qwen Community 1.0",
+  "licenseUrl": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next/blob/main/LICENSE",
+  "parameterSize": "180B",
+  "approxSizeBytes": 81961823936,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "upstream": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next",
+  "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.retrieval-first",
+    "prompt.workspace-gestalt",
+    "prompt.derive-by-execution"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 1,
+      "topP": 0.95,
+      "topK": 20,
+      "minP": 0,
+      "presencePenalty": 0,
+      "repetitionPenalty": 1,
+      "maxTokens": 12288
+    },
+    "samplingWhenThinking": {
+      "temperature": 1,
+      "topP": 0.95
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 4096,
+      "templateKwargs": {
+        "reasoning_effort": "xhigh"
+      }
+    },
+    "promptTags": {
+      "enableThinkingTag": "/think",
+      "disableThinkingTag": "/no_think"
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 12288
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096,
+          "templateKwargs": {
+            "reasoning_effort": "xhigh"
+          }
+        }
+      },
+      "thinking-deep": {
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096,
+          "templateKwargs": {
+            "reasoning_effort": "xhigh"
+          }
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096,
+          "templateKwargs": {
+            "reasoning_effort": "xhigh"
+          }
+        }
+      },
+      "thinking-precise": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 0,
+          "repetitionPenalty": 1,
+          "seed": 0,
+          "maxTokens": 6144
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096,
+          "templateKwargs": {
+            "reasoning_effort": "medium"
+          }
+        }
+      },
+      "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,
+          "topP": 0.95,
+          "topK": 40,
+          "minP": 0,
+          "presencePenalty": 1.5,
+          "repetitionPenalty": 1,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      }
+    },
+    "engine": {
+      "llamaCpp": {
+        "loadMode": "auto",
+        "lazyMode": "auto",
+        "cacheReuse": 0,
+        "flashAttn": "on",
+        "batchSize": 512,
+        "ubatchSize": 512,
+        "contextSize": 65536
+      }
+    }
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "unsloth/Qwen3.8-Flash-Next-GGUF",
+    "shardsDir": "UD-IQ3_XXS",
+    "quantization": "UD-IQ3_XXS",
+    "residentBytes": 53150661120,
+    "mmprojFilename": "mmproj-F16.gguf"
+  }
+}

package/authoring/chat-models/qwen3.8-flash-next-iq4.json

@@ -0,0 +1,198 @@
+{
+  "id": "qwen3.8-flash-next-iq4",
+  "name": "Qwen 3.8 Flash Next (180B-A6B, IQ4_XS)",
+  "description": "Qwen's experimental preview of the architecture behind Qwen4: a 125B-parameter MoE with 6B active per token, plus a 51B n-gram embedding table designed for cheap offload. This higher-fidelity Unsloth UD-IQ4_XS build is a 94 GB split GGUF whose 26.8 GiB n-gram table can be demand-paged. It needs roughly 60.4 GiB for resident weights before KV cache and runtime overhead, so 96 GB system RAM is recommended; a 64 GB system with a 32 GB discrete GPU is possible but leaves almost no operating headroom.",
+  "tags": [
+    "qwen",
+    "agentic",
+    "coding",
+    "multimodal",
+    "vision",
+    "video",
+    "tools",
+    "reasoning",
+    "long-context",
+    "moe",
+    "ngram-embedding",
+    "large",
+    "experimental"
+  ],
+  "category": "general",
+  "maintainer": {
+    "name": "Unsloth",
+    "url": "https://huggingface.co/unsloth/Qwen3.8-Flash-Next-GGUF"
+  },
+  "maker": {
+    "name": "Qwen",
+    "url": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next"
+  },
+  "version": "1.0.0",
+  "versionMinGezelVersion": "1.26248",
+  "updatedAt": "2026-09-06T00:00:00Z",
+  "license": "qwen-community-1.0",
+  "licenseClass": "custom-restricted",
+  "licenseShortName": "Qwen Community 1.0",
+  "licenseUrl": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next/blob/main/LICENSE",
+  "parameterSize": "180B",
+  "approxSizeBytes": 93682584224,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "upstream": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next",
+  "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.retrieval-first",
+    "prompt.workspace-gestalt",
+    "prompt.derive-by-execution"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 1,
+      "topP": 0.95,
+      "topK": 20,
+      "minP": 0,
+      "presencePenalty": 0,
+      "repetitionPenalty": 1,
+      "maxTokens": 12288
+    },
+    "samplingWhenThinking": {
+      "temperature": 1,
+      "topP": 0.95
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 4096,
+      "templateKwargs": {
+        "reasoning_effort": "xhigh"
+      }
+    },
+    "promptTags": {
+      "enableThinkingTag": "/think",
+      "disableThinkingTag": "/no_think"
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 12288
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096,
+          "templateKwargs": {
+            "reasoning_effort": "xhigh"
+          }
+        }
+      },
+      "thinking-deep": {
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096,
+          "templateKwargs": {
+            "reasoning_effort": "xhigh"
+          }
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096,
+          "templateKwargs": {
+            "reasoning_effort": "xhigh"
+          }
+        }
+      },
+      "thinking-precise": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 0,
+          "repetitionPenalty": 1,
+          "seed": 0,
+          "maxTokens": 6144
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096,
+          "templateKwargs": {
+            "reasoning_effort": "medium"
+          }
+        }
+      },
+      "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,
+          "topP": 0.95,
+          "topK": 40,
+          "minP": 0,
+          "presencePenalty": 1.5,
+          "repetitionPenalty": 1,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      }
+    },
+    "engine": {
+      "llamaCpp": {
+        "loadMode": "auto",
+        "lazyMode": "auto",
+        "cacheReuse": 0,
+        "flashAttn": "on",
+        "batchSize": 512,
+        "ubatchSize": 512,
+        "contextSize": 65536
+      }
+    }
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "unsloth/Qwen3.8-Flash-Next-GGUF",
+    "shardsDir": "UD-IQ4_XS",
+    "quantization": "UD-IQ4_XS",
+    "residentBytes": 64871421440,
+    "mmprojFilename": "mmproj-F16.gguf"
+  }
+}

package/authoring/gstack/wave.json

@@ -1,7 +1,7 @@
 {
   "schemaVersion": 1,
-  "version": "2.0.7",
-  "releasedAt": "2026-09-04T13:46:25Z",
+  "version": "2.0.8",
+  "releasedAt": "2026-09-05T03:30:00Z",
   "sourceRevision": "a3259400a366593e0c909dd9ac3e59752efd2488",
   "basedOn": {
     "name": "gstack",

package/data/chat-models/gl/glm-5.3-flash-320b-q2/manifest.json

@@ -0,0 +1,137 @@
+{
+  "schemaVersion": 1,
+  "kind": "chat-model",
+  "id": "glm-5.3-flash-320b-q2",
+  "name": "GLM 5.3 Flash (320B-A18B, IQ2_XXS)",
+  "description": "antirez's routed 2-bit GGUF of Z.ai's GLM 5.3 Flash — a natively multimodal 320B-parameter Mixture-of-Experts with 18B active parameters, a 1M-token context, recurrent KDA layers, sparse DSA attention, hyper-connections, and a built-in MTP draft block. The Q2 file is about 90 GiB and targets a 128 GB Mac or DGX Spark with a modest initial context. Its separately packaged vision encoder adds image understanding to the same text model when Gezel launches ds4 with `--vision`.",
+  "tags": [
+    "glm",
+    "zai",
+    "agentic",
+    "coding",
+    "multimodal",
+    "vision",
+    "reasoning",
+    "moe",
+    "tools",
+    "2-bit",
+    "iq2",
+    "long-context",
+    "large"
+  ],
+  "category": "reasoning",
+  "maintainer": {
+    "name": "antirez",
+    "url": "https://huggingface.co/antirez/glm-5.3-flash-gguf"
+  },
+  "maker": {
+    "name": "Z.ai",
+    "url": "https://huggingface.co/zai-org/GLM-5.3-Flash"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-09-06T00:00:00Z",
+  "license": "MIT",
+  "licenseClass": "open",
+  "licenseShortName": "MIT",
+  "parameterSize": "320B",
+  "approxSizeBytes": 97633097344,
+  "supportsTools": true,
+  "contextWindow": 1048576,
+  "upstream": "https://huggingface.co/zai-org/GLM-5.3-Flash",
+  "style": {
+    "family": "glm",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.tool-cookbook-condensed",
+    "prompt.meester-build-prelude",
+    "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
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "thinking-precise": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "seed": 0,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "instruct": {
+        "sampling": {
+          "temperature": 0.7,
+          "topP": 1,
+          "maxTokens": 4096
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      },
+      "creative": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      }
+    }
+  },
+  "ds4": {
+    "huggingfaceRepo": "antirez/glm-5.3-flash-gguf",
+    "revision": "b2fa29d7a6b410db11221c904973967b80b760f5",
+    "filename": "GLM-5.3-Flash-Q2.gguf",
+    "sha256": "e81fd6241c6e55a64e1e14e47a3eab61a173fa8d7e4b5c1d1848827119705b32",
+    "approxSizeBytes": 96505816384,
+    "quantization": "IQ2_XXS/Q2_K",
+    "visionEncoder": {
+      "filename": "GLM-5.3-Flash-Vision-Encoder.gguf",
+      "sha256": "ae23e14c6979e889051b2e4a39351abcdafb161e18e606fae4d8c40095a4bf3a",
+      "sizeBytes": 1127280960
+    },
+    "residentBytes": 100865554432,
+    "ssdStreaming": false,
+    "maxLaunchCtx": 32768
+  }
+}

package/data/chat-models/gl/glm-5.3-flash-320b-q2/versions/1.0.0/manifest.json

@@ -0,0 +1,23 @@
+{
+  "schemaVersion": 1,
+  "version": "1.0.0",
+  "releasedAt": "2026-09-06T00:00:00Z",
+  "minGezelVersion": "1.26248",
+  "approxSizeBytes": 97633097344,
+  "ds4": {
+    "huggingfaceRepo": "antirez/glm-5.3-flash-gguf",
+    "revision": "b2fa29d7a6b410db11221c904973967b80b760f5",
+    "filename": "GLM-5.3-Flash-Q2.gguf",
+    "sha256": "e81fd6241c6e55a64e1e14e47a3eab61a173fa8d7e4b5c1d1848827119705b32",
+    "approxSizeBytes": 96505816384,
+    "quantization": "IQ2_XXS/Q2_K",
+    "visionEncoder": {
+      "filename": "GLM-5.3-Flash-Vision-Encoder.gguf",
+      "sha256": "ae23e14c6979e889051b2e4a39351abcdafb161e18e606fae4d8c40095a4bf3a",
+      "sizeBytes": 1127280960
+    },
+    "residentBytes": 100865554432,
+    "ssdStreaming": false,
+    "maxLaunchCtx": 32768
+  }
+}