Skip to content

Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.

4.6 KB
Raw Blame
[
  {
    "id": "qwen3.5-4b",
    "name": "Qwen 3.5 4B",
    "description": "Newest small Qwen. Tops independent tool-calling tests for its size and stays fast on modest GPUs.",
    "hf_repo": "unsloth/Qwen3.5-4B-GGUF",
    "hf_filename": "Qwen3.5-4B-Q4_K_M.gguf",
    "size_bytes": 2740937888,
    "quant": "Q4_K_M",
    "ctx": 16384,
    "vram_estimate_mb": 4600,
    "ngl_suggestion": 99,
    "tags": [
      "fast",
      "4b"
    ],
    "sampling": {
      "temperature": 0.7,
      "top_p": 0.8,
      "top_k": 20,
      "presence_penalty": 1.5
    }
  },
  {
    "id": "qwen3.5-9b",
    "name": "Qwen 3.5 9B",
    "description": "Best balance of reliability and speed on an 8 GB+ GPU. Noticeably steadier than 4B over long conversations.",
    "hf_repo": "unsloth/Qwen3.5-9B-GGUF",
    "hf_filename": "Qwen3.5-9B-Q4_K_M.gguf",
    "size_bytes": 5680522464,
    "quant": "Q4_K_M",
    "ctx": 16384,
    "vram_estimate_mb": 7600,
    "ngl_suggestion": 99,
    "tags": [
      "balanced",
      "9b"
    ],
    "sampling": {
      "temperature": 0.7,
      "top_p": 0.8,
      "top_k": 20,
      "presence_penalty": 1.5
    }
  },
  {
    "id": "qwen3-4b",
    "name": "Qwen 3 4B Instruct",
    "description": "Latest Qwen generation. Fast, excellent tool-calling for its size, the recommended starting point.",
    "hf_repo": "unsloth/Qwen3-4B-Instruct-2507-GGUF",
    "hf_filename": "Qwen3-4B-Instruct-2507-Q4_K_M.gguf",
    "size_bytes": 2497281120,
    "quant": "Q4_K_M",
    "ctx": 16384,
    "vram_estimate_mb": 4200,
    "ngl_suggestion": 99,
    "tags": [
      "fast",
      "4b",
      "default"
    ]
  },
  {
    "id": "qwen3-8b",
    "name": "Qwen 3 8B",
    "description": "Great quality/speed balance with hybrid thinking. Fits in 8 GB VRAM.",
    "hf_repo": "unsloth/Qwen3-8B-GGUF",
    "hf_filename": "Qwen3-8B-Q4_K_M.gguf",
    "size_bytes": 5027784512,
    "quant": "Q4_K_M",
    "ctx": 16384,
    "vram_estimate_mb": 7000,
    "ngl_suggestion": 99,
    "tags": [
      "balanced",
      "8b",
      "thinking"
    ]
  },
  {
    "id": "qwen3-14b",
    "name": "Qwen 3 14B",
    "description": "High quality with hybrid thinking (reasons before answering when useful). Needs ~11 GB VRAM.",
    "hf_repo": "unsloth/Qwen3-14B-GGUF",
    "hf_filename": "Qwen3-14B-Q4_K_M.gguf",
    "size_bytes": 9001753984,
    "quant": "Q4_K_M",
    "ctx": 16384,
    "vram_estimate_mb": 11000,
    "ngl_suggestion": 99,
    "tags": [
      "quality",
      "14b",
      "thinking"
    ]
  },
  {
    "id": "gpt-oss-20b",
    "name": "GPT-OSS 20B (OpenAI)",
    "description": "OpenAI's open-weight MoE (3.6B active params, fast for its size). Top-tier tool calling and reasoning. Best with 16 GB VRAM.",
    "hf_repo": "ggml-org/gpt-oss-20b-GGUF",
    "hf_filename": "gpt-oss-20b-MXFP4.gguf",
    "size_bytes": 12109566624,
    "quant": "MXFP4",
    "ctx": 16384,
    "vram_estimate_mb": 13500,
    "ngl_suggestion": 99,
    "tags": [
      "quality",
      "20b",
      "thinking",
      "moe"
    ]
  },
  {
    "id": "qwen3-vl-4b",
    "name": "Qwen 3 VL 4B (vision, light)",
    "description": "Small vision model, sees attached images while fitting in ~5 GB VRAM. Pick this over the 8B on smaller GPUs.",
    "hf_repo": "unsloth/Qwen3-VL-4B-Instruct-GGUF",
    "hf_filename": "Qwen3-VL-4B-Instruct-Q4_K_M.gguf",
    "hf_mmproj_filename": "mmproj-F16.gguf",
    "mmproj_size_bytes": 836180640,
    "size_bytes": 2497282336,
    "quant": "Q4_K_M",
    "ctx": 16384,
    "vram_estimate_mb": 4600,
    "ngl_suggestion": 99,
    "tags": [
      "vision",
      "4b",
      "fast"
    ]
  },
  {
    "id": "qwen3-vl-8b",
    "name": "Qwen 3 VL 8B (vision)",
    "description": "Sees images, the agent can look at attached pictures before picking a workflow. Needs ~8 GB VRAM.",
    "hf_repo": "unsloth/Qwen3-VL-8B-Instruct-GGUF",
    "hf_filename": "Qwen3-VL-8B-Instruct-Q4_K_M.gguf",
    "hf_mmproj_filename": "mmproj-F16.gguf",
    "mmproj_size_bytes": 1159030336,
    "size_bytes": 5027785568,
    "quant": "Q4_K_M",
    "ctx": 16384,
    "vram_estimate_mb": 7800,
    "ngl_suggestion": 99,
    "tags": [
      "vision",
      "8b"
    ]
  },
  {
    "id": "cadquery-coder-7b",
    "name": "CadQuery Coder 7B (CAD-specialized)",
    "description": "Qwen2.5-Coder-7B fine-tuned on CadQuery generation, best pick for the Text to CAD node. Community model, CC-BY-NC-SA license.",
    "hf_repo": "yuvit-batra/qwen2.5-coder-7b-cadquery-gguf",
    "hf_filename": "qwen2.5-coder-7b-cadquery-Q4_K_M.gguf",
    "size_bytes": 4680000000,
    "quant": "Q4_K_M",
    "ctx": 16384,
    "vram_estimate_mb": 6200,
    "ngl_suggestion": 99,
    "tags": [
      "code",
      "cad",
      "7b"
    ]
  }
]