@bendyline/gilde 0.1.17 → 0.1.19

package/authoring/chat-models/laguna-s-2.1-118b-q8.json

@@ -0,0 +1,127 @@
+{
+  "id": "laguna-s-2.1-118b-q8",
+  "name": "Laguna S 2.1 (118B-A8B, Q8)",
+  "description": "Poolside's Laguna S 2.1, a 118B-parameter Mixture-of-Experts model with about 8B active parameters per token, built for agentic coding and long-horizon software work. It supports native tool calling and interleaved reasoning between tool calls. The official Q8_0 GGUF is about 128GB on disk, while the community MLX 8-bit conversion is about 125GB and runs natively on Apple Silicon. Both target larger-memory workstations; Q6 is the practical Laguna tier for a 128GB Mac. Gezel defaults to Poolside's quality-recommended 256K context window; the underlying checkpoint can be extended to 1M context with explicit YaRN overrides.",
+  "tags": [
+    "poolside",
+    "coding",
+    "agentic",
+    "tools",
+    "reasoning",
+    "long-context",
+    "moe",
+    "large",
+    "open-weights",
+    "apple-silicon"
+  ],
+  "category": "coding",
+  "maintainer": {
+    "name": "Poolside",
+    "url": "https://huggingface.co/poolside/Laguna-S-2.1"
+  },
+  "version": "1.0.1",
+  "updatedAt": "2026-07-28T00:00:00Z",
+  "license": "OpenMDW-1.1",
+  "licenseClass": "open",
+  "licenseShortName": "OpenMDW 1.1",
+  "licenseUrl": "https://huggingface.co/poolside/Laguna-S-2.1/blob/main/LICENSE.md",
+  "parameterSize": "118B",
+  "approxSizeBytes": 128750823168,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "upstream": "https://huggingface.co/poolside/Laguna-S-2.1",
+  "style": {
+    "family": "other",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.preamble-folding",
+    "turn.ramble-detection",
+    "tools.mlx-grammar"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 0.7,
+      "topP": 0.95,
+      "maxTokens": 8192
+    },
+    "samplingWhenThinking": {
+      "temperature": 0.7,
+      "topP": 0.95
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 4096
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 0.7,
+          "topP": 0.95,
+          "maxTokens": 12288
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 6144
+        }
+      },
+      "thinking-precise": {
+        "sampling": {
+          "temperature": 0.4,
+          "topP": 0.9,
+          "seed": 0,
+          "maxTokens": 6144
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "instruct": {
+        "sampling": {
+          "temperature": 0.7,
+          "topP": 0.95,
+          "maxTokens": 4096
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      },
+      "creative": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      }
+    }
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "poolside/Laguna-S-2.1-GGUF",
+    "filename": "laguna-s-2.1-Q8_0.gguf",
+    "quantization": "Q8_0",
+    "residentBytes": 154500987802
+  },
+  "mlx": {
+    "huggingfaceRepo": "Vontra/Laguna-S-2.1-MLX-8bit",
+    "quantization": "8bit",
+    "residentBytes": 125000000000
+  }
+}

package/authoring/chat-models/lfm2.5-2.6b-q4.json

@@ -0,0 +1,39 @@
+{
+  "id": "lfm2.5-2.6b-q4",
+  "name": "LFM2.5 (2.6B)",
+  "description": "Liquid AI's LFM2.5 at 2.6B parameters. A hybrid convolution-attention design built for on-device speed — light on memory, quick even on modest hardware, with integrated reasoning, native tool calling, and a 128K context window.",
+  "tags": [
+    "liquid-ai",
+    "small",
+    "tools"
+  ],
+  "category": "general",
+  "maintainer": {
+    "name": "Liquid AI",
+    "url": "https://huggingface.co/LiquidAI/LFM2.5-2.6B"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-08-04T00:00:00Z",
+  "license": "LFM Open License v1.0",
+  "licenseClass": "custom-restricted",
+  "licenseShortName": "LFM Open 1.0",
+  "parameterSize": "2.6B",
+  "approxSizeBytes": 1674454848,
+  "supportsTools": true,
+  "contextWindow": 128000,
+  "upstream": "https://huggingface.co/LiquidAI/LFM2.5-2.6B",
+  "style": {
+    "family": "other",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "LiquidAI/LFM2.5-2.6B-GGUF",
+    "filename": "LFM2.5-2.6B-Q4_K_M.gguf",
+    "quantization": "Q4_K_M"
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/LFM2.5-2.6B-4bit",
+    "quantization": "4bit"
+  }
+}

package/authoring/chat-models/llama3.2-3b-q4.json

@@ -0,0 +1,42 @@
+{
+  "id": "llama3.2-3b-q4",
+  "name": "Llama 3.2 (3B)",
+  "description": "Meta's 3B chat model. Small enough to run on modest hardware, sharp enough to be a useful default.",
+  "tags": [
+    "meta",
+    "small"
+  ],
+  "category": "general",
+  "maintainer": {
+    "name": "Meta",
+    "url": "https://ollama.com/library/llama3.2"
+  },
+  "version": "1.1.0",
+  "updatedAt": "2026-04-26T00:00:00Z",
+  "license": "Llama 3.2 Community License",
+  "parameterSize": "3B",
+  "approxSizeBytes": 2019377696,
+  "supportsTools": true,
+  "contextWindow": 128000,
+  "upstream": "https://ollama.com/library/llama3.2",
+  "style": {
+    "family": "llama",
+    "reasoningFormat": "none",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "fabrication.detect-past-tense-no-tools"
+  ],
+  "ollama": {
+    "tag": "llama3.2"
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "bartowski/Llama-3.2-3B-Instruct-GGUF",
+    "filename": "Llama-3.2-3B-Instruct-Q4_K_M.gguf",
+    "quantization": "Q4_K_M"
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/Llama-3.2-3B-Instruct-4bit",
+    "quantization": "4bit"
+  }
+}

package/authoring/chat-models/mistral-7b-q4.json

@@ -0,0 +1,53 @@
+{
+  "id": "mistral-7b-q4",
+  "name": "Mistral 7B Instruct",
+  "description": "Well-rounded 7B instruct model from Mistral AI. Good general-purpose default.",
+  "tags": [
+    "mistral",
+    "tools"
+  ],
+  "category": "general",
+  "maintainer": {
+    "name": "Mistral AI",
+    "url": "https://ollama.com/library/mistral"
+  },
+  "version": "1.1.0",
+  "updatedAt": "2026-04-26T00:00:00Z",
+  "license": "Apache-2.0",
+  "parameterSize": "7B",
+  "approxSizeBytes": 4372812000,
+  "supportsTools": true,
+  "contextWindow": 32768,
+  "upstream": "https://ollama.com/library/mistral",
+  "style": {
+    "family": "mistral",
+    "reasoningFormat": "none",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "fabrication.detect-past-tense-no-tools",
+    "mcp.compact-tool-schemas"
+  ],
+  "tuning": {
+    "engine": {
+      "llamaCpp": {
+        "chatTemplate": "{{ bos_token }}\n{% if tools %}\n{{ '[AVAILABLE_TOOLS] ' + (tools | tojson) + ' [/AVAILABLE_TOOLS]' }}\n{% endif %}\n{% for message in messages %}\n{% if message['role'] == 'system' %}\n{{ message['content'] }}\n{% elif message['role'] == 'user' %}\n{{ '[INST] ' + message['content'] + ' [/INST]' }}\n{% elif message['role'] == 'assistant' %}\n{% if message.get('tool_calls') %}\n{{ '[TOOL_CALLS] ' + (message['tool_calls'] | tojson) }}\n{% else %}\n{{ message['content'] + eos_token }}\n{% endif %}\n{% elif message['role'] == 'tool' %}\n{{ '[TOOL_RESULTS] ' + message['content'] + ' [/TOOL_RESULTS]' }}\n{% endif %}\n{% endfor %}"
+      }
+    },
+    "sampling": {
+      "maxTokens": 8192
+    }
+  },
+  "ollama": {
+    "tag": "mistral:7b"
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "bartowski/Mistral-7B-Instruct-v0.3-GGUF",
+    "filename": "Mistral-7B-Instruct-v0.3-Q4_K_M.gguf",
+    "quantization": "Q4_K_M"
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/Mistral-7B-Instruct-v0.3-4bit",
+    "quantization": "4bit"
+  }
+}

package/authoring/chat-models/mistral-medium-3.5-128b-q4.json

@@ -0,0 +1,57 @@
+{
+  "id": "mistral-medium-3.5-128b-q4",
+  "name": "Mistral Medium 3.5 (128B)",
+  "description": "Mistral AI's frontier dense 128B model with configurable reasoning (instant or chain-of-thought via `reasoning_effort`). Native function-calling, 256K context. Workstation-tier on-device — needs ~80GB of usable memory for Q4_K_M; an MLX 4-bit build is available for Apple Silicon Mac Studio-class hardware. License is a Modified MIT with a revenue carve-out — free for commercial and non-commercial use unless the licensee company exceeds Mistral's revenue threshold.",
+  "tags": [
+    "mistral",
+    "reasoning",
+    "tools",
+    "large",
+    "revenue-limited"
+  ],
+  "category": "reasoning",
+  "maintainer": {
+    "name": "Mistral AI",
+    "url": "https://huggingface.co/mistralai/Mistral-Medium-3.5-128B"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-05-22T00:00:00Z",
+  "license": "MIT-Modified",
+  "parameterSize": "128B",
+  "approxSizeBytes": 72000000000,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "upstream": "https://huggingface.co/mistralai/Mistral-Medium-3.5-128B",
+  "style": {
+    "family": "mistral",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.preamble-folding"
+  ],
+  "llamaCpp": {
+    "huggingfaceRepo": "unsloth/Mistral-Medium-3.5-128B-GGUF",
+    "shardsDir": "Q4_K_M",
+    "quantization": "Q4_K_M"
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/Mistral-Medium-3.5-128B-4bit",
+    "quantization": "4bit"
+  },
+  "tuning": {
+    "sampling": {
+      "temperature": 0.7,
+      "topP": 0.95,
+      "topK": 64,
+      "repetitionPenalty": 1.05,
+      "repetitionContext": 20,
+      "maxTokens": 8192
+    },
+    "reasoning": {
+      "thinkingBudget": 4096
+    }
+  }
+}

package/authoring/chat-models/muse-glimmer-30b-q4.json

@@ -0,0 +1,126 @@
+{
+  "id": "muse-glimmer-30b-q4",
+  "name": "Muse Glimmer (30B, Q4)",
+  "description": "Meta Superintelligence Labs' Muse family, distilled from the larger Muse Spark down to 30B: a dense agentic model (29.6B language model plus a ~1.8B ViT-G/14 perception encoder) built for long-horizon tool use on consumer hardware. A new architecture rather than a Llama derivative — its own chat format and tool-call grammar. Multimodal (text + image), 131K native context, native tool calling, and a reasoning strength dial from low to xhigh. Meta's own calibrated K-quant build, sized to fit 24GB of VRAM.",
+  "tags": [
+    "meta",
+    "agentic",
+    "multimodal",
+    "vision",
+    "tools",
+    "long-context",
+    "reasoning"
+  ],
+  "category": "general",
+  "maintainer": {
+    "name": "Meta Superintelligence Labs",
+    "url": "https://huggingface.co/meta-models/Muse-Glimmer-30B"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-08-10T00:00:00Z",
+  "license": "Apache-2.0",
+  "licenseClass": "open",
+  "licenseShortName": "Apache 2.0",
+  "parameterSize": "30B",
+  "approxSizeBytes": 16756681056,
+  "supportsTools": true,
+  "contextWindow": 131072,
+  "upstream": "https://huggingface.co/meta-models/Muse-Glimmer-30B",
+  "minGezelVersion": "1.99999",
+  "style": {
+    "family": "muse",
+    "reasoningFormat": "channel",
+    "toolCallFormat": "function-call"
+  },
+  "tuning": {
+    "sampling": {
+      "temperature": 1,
+      "topP": 0.95,
+      "topK": 64,
+      "maxTokens": 8192
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "templateKwargs": {
+        "reasoning_strength": "high"
+      }
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "topK": 64,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "templateKwargs": {
+            "reasoning_strength": "high"
+          }
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "topK": 64,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "templateKwargs": {
+            "reasoning_strength": "xhigh"
+          }
+        }
+      },
+      "thinking-precise": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.9,
+          "topK": 64,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "templateKwargs": {
+            "reasoning_strength": "high"
+          }
+        }
+      },
+      "instruct": {
+        "sampling": {
+          "temperature": 0.7,
+          "topP": 0.95,
+          "topK": 64,
+          "maxTokens": 4096
+        },
+        "reasoning": {
+          "templateKwargs": {
+            "reasoning_strength": "low"
+          }
+        }
+      },
+      "creative": {
+        "sampling": {
+          "temperature": 1.1,
+          "topP": 0.95,
+          "topK": 64,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "templateKwargs": {
+            "reasoning_strength": "low"
+          }
+        }
+      }
+    }
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "meta-models/Muse-Glimmer-30B-GGUF",
+    "filename": "muse-glimmer-30B-kquant-17gb.gguf",
+    "quantization": "K-Quant-17GB",
+    "mmprojFilename": "mmproj-kquant.gguf",
+    "residentBytes": 21416116224
+  }
+}

package/authoring/chat-models/nemotron3-nano-30b-q4.json

@@ -0,0 +1,49 @@
+{
+  "id": "nemotron3-nano-30b-q4",
+  "name": "Nemotron-3 Nano (30B, MoE Reasoning)",
+  "description": "NVIDIA's Nemotron-3 Nano Omni — 30B-parameter hybrid Mamba/Transformer Mixture-of-Experts (~3B active per token) tuned for reasoning. Native function-calling, 128K context, multimodal (text + image + audio + video).",
+  "tags": [
+    "nvidia",
+    "reasoning",
+    "moe",
+    "hybrid",
+    "tools",
+    "multimodal",
+    "vision",
+    "audio",
+    "video"
+  ],
+  "category": "reasoning",
+  "maintainer": {
+    "name": "NVIDIA",
+    "url": "https://huggingface.co/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-05-17T00:00:00Z",
+  "license": "NVIDIA Open Model License",
+  "parameterSize": "30B",
+  "approxSizeBytes": 24515129536,
+  "supportsTools": true,
+  "contextWindow": 131072,
+  "upstream": "https://huggingface.co/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning",
+  "style": {
+    "family": "nemotron",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.preamble-folding"
+  ],
+  "llamaCpp": {
+    "huggingfaceRepo": "lmstudio-community/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-GGUF",
+    "filename": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning-Q4_K_M.gguf",
+    "quantization": "Q4_K_M",
+    "mmprojFilename": "mmproj-Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16.gguf"
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-nvfp4",
+    "quantization": "nvfp4"
+  }
+}

package/authoring/chat-models/nemotron3-super-120b-q4.json

@@ -0,0 +1,41 @@
+{
+  "id": "nemotron3-super-120b-q4",
+  "name": "Nemotron-3 Super (120B-A12B)",
+  "description": "NVIDIA's flagship Nemotron-3 Super — 120B-parameter hybrid Mamba/Transformer Mixture-of-Experts (~12B active per token). Native function-calling, 128K context. Frontier-tier on-device reasoning for workstations and Mac Studio-class hardware — needs ~100GB of usable memory for Q4_K_M. MLX path is held until mlx-community publishes a build that doesn't require trust_remote_code; for now this is llama.cpp-only.",
+  "tags": [
+    "nvidia",
+    "reasoning",
+    "moe",
+    "hybrid",
+    "tools",
+    "large"
+  ],
+  "category": "reasoning",
+  "maintainer": {
+    "name": "NVIDIA",
+    "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-05-17T00:00:00Z",
+  "license": "NVIDIA Open Model License",
+  "parameterSize": "120B",
+  "approxSizeBytes": 86051079584,
+  "supportsTools": true,
+  "contextWindow": 131072,
+  "upstream": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
+  "style": {
+    "family": "nemotron",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.preamble-folding"
+  ],
+  "llamaCpp": {
+    "huggingfaceRepo": "lmstudio-community/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF",
+    "shardsPrefix": "NVIDIA-Nemotron-3-Super-120B-A12B-Q4_K_M-",
+    "quantization": "Q4_K_M"
+  }
+}

package/authoring/chat-models/ornith-35b-q4.json

@@ -0,0 +1,145 @@
+{
+  "id": "ornith-35b-q4",
+  "name": "Ornith 1.0 (35B, Q4)",
+  "description": "DeepReinforce's Ornith 1.0 at 35B total parameters (Qwen 3.5 MoE architecture), an agentic-coding model trained with a self-improving reinforcement-learning framework — state-of-the-art on SWE-Bench and Terminal-Bench. Native <think> reasoning, OpenAI-style tool calling, 256K context. Q4_K_M quantization for a ~21GB footprint. MIT licensed.",
+  "tags": [
+    "deepreinforce",
+    "coding",
+    "agentic",
+    "tools",
+    "reasoning",
+    "long-context",
+    "moe"
+  ],
+  "category": "coding",
+  "maintainer": {
+    "name": "DeepReinforce",
+    "url": "https://huggingface.co/deepreinforce-ai"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-07-01T00:00:00Z",
+  "license": "MIT",
+  "licenseClass": "open",
+  "licenseShortName": "MIT",
+  "parameterSize": "35B",
+  "approxSizeBytes": 21200000000,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "kvBytesPerTokenF16": 20480,
+  "kvFixedBytesF16": 64389120,
+  "upstream": "https://huggingface.co/deepreinforce-ai/Ornith-1.0-35B",
+  "style": {
+    "family": "qwen",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.tool-cookbook-condensed",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.ollama-num-predict-bumped",
+    "turn.preamble-folding",
+    "turn.ramble-detection"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 0.6,
+      "topP": 0.95,
+      "topK": 20,
+      "minP": 0,
+      "maxTokens": 8192
+    },
+    "samplingWhenThinking": {
+      "temperature": 0.6,
+      "topP": 0.95
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 4096
+    },
+    "promptTags": {
+      "enableThinkingTag": "/think",
+      "disableThinkingTag": "/no_think"
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 0.6,
+          "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
+        }
+      }
+    }
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "deepreinforce-ai/Ornith-1.0-35B-GGUF",
+    "filename": "ornith-1.0-35b-Q4_K_M.gguf",
+    "quantization": "Q4_K_M",
+    "residentBytes": 25400109312
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/Ornith-1.0-35B-4bit",
+    "quantization": "4bit",
+    "residentBytes": 26557921365
+  }
+}