cstr/bge-m3-gguf overview

{
  "metadata": {},
  "card_data": {
    "license": "mit",
    "language": [
      "multilingual"
    ],
    "tags": [
      "embeddings",
      "gguf",
      "ggml",
      "text-embeddings",
      "xlm-r",
      "crispembed"
    ],
    "pipeline_tag": "feature-extraction",
    "base_model": "BAAI/bge-m3",
    "frontmatter": {
      "license": "mit",
      "language": "[multilingual]",
      "tags": "[embeddings, gguf, ggml, text-embeddings, xlm-r, crispembed]",
      "pipeline_tag": "feature-extraction",
      "base_model": "BAAI/bge-m3"
    },
    "hero_image_url": "",
    "summary": "GGUF format of BAAI/bge-m3 for use with CrispEmbed. BGE-M3. Dense + sparse + ColBERT multi-vector retrieval in one model. 100+ languages, 8192 context.",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nlicense: mit\nlanguage: [multilingual]\ntags: [embeddings, gguf, ggml, text-embeddings, xlm-r, crispembed]\npipeline_tag: feature-extraction\nbase_model: BAAI/bge-m3\n---\n\n# bge-m3 GGUF\n\nGGUF format of [BAAI/bge-m3](https://huggingface.co/BAAI/bge-m3) for use with [CrispEmbed](https://github.com/CrispStrobe/CrispEmbed).\n\nBGE-M3. Dense + sparse + ColBERT multi-vector retrieval in one model. 100+ languages, 8192 context.\n\n## Files\n\n| File | Quantization | Size |\n|------|-------------|------|\n| [bge-m3-q4_k.gguf](https://huggingface.co/cstr/bge-m3-GGUF/resolve/main/bge-m3-q4_k.gguf) | Q4_K | 438 MB |\n| [bge-m3-q8_0.gguf](https://huggingface.co/cstr/bge-m3-GGUF/resolve/main/bge-m3-q8_0.gguf) | Q8_0 | 583 MB |\n| [bge-m3.gguf](https://huggingface.co/cstr/bge-m3-GGUF/resolve/main/bge-m3.gguf) | F32 | 2175 MB |\n\n\n## Quick Start\n\n```bash\n# Download\nhuggingface-cli download cstr/bge-m3-GGUF bge-m3-q4_k.gguf --local-dir .\n\n# Run with CrispEmbed\n./crispembed -m bge-m3-q4_k.gguf \"Hello world\"\n\n# Or with auto-download\n./crispembed -m bge-m3 \"Hello world\"\n```\n\n## Model Details\n\n| Property | Value |\n|----------|-------|\n| Architecture | XLM-R |\n| Parameters | 568M |\n| Embedding Dimension | 1024 |\n| Layers | 24 |\n| Pooling | mean |\n| Tokenizer | SentencePiece |\n| Base Model | [BAAI/bge-m3](https://huggingface.co/BAAI/bge-m3) |\n\n## Verification\n\nVerified bit-identical to HuggingFace sentence-transformers (cosine similarity >= 0.999 on test texts).\n\n## Usage with CrispEmbed\n\nCrispEmbed is a lightweight C/C++ text embedding inference engine using ggml.\nNo Python runtime, no ONNX. Supports BERT, XLM-R, Qwen3, and Gemma3 architectures.\n\n```bash\n# Build CrispEmbed\ngit clone https://github.com/CrispStrobe/CrispEmbed\ncd CrispEmbed\ncmake -S . -B build && cmake --build build -j\n\n# Encode\n./build/crispembed -m bge-m3-q4_k.gguf \"query text\"\n\n# Server mode\n./build/crispembed-server -m bge-m3-q4_k.gguf --port 8080\ncurl -X POST http://localhost:8080/v1/embeddings \\\n    -d '{\"input\": [\"Hello world\"], \"model\": \"bge-m3\"}'\n```\n\n## Credits\n\n- Original model: [BAAI/bge-m3](https://huggingface.co/BAAI/bge-m3)\n- Inference engine: [CrispEmbed](https://github.com/CrispStrobe/CrispEmbed) (ggml-based)\n- Conversion: `convert-bert-embed-to-gguf.py`\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "embeddings",
    "ggml",
    "text-embeddings",
    "xlm-r",
    "crispembed",
    "feature-extraction",
    "multilingual",
    "base_model:BAAI/bge-m3",
    "base_model:quantized:BAAI/bge-m3",
    "license:mit",
    "region:us"
  ],
  "likes": 0,
  "downloads": 185,
  "gated": false,
  "private": false,
  "last_modified": "2026-04-17T07:19:29.000Z",
  "created_at": "2026-04-16T22:44:14.000Z",
  "pipeline_tag": "feature-extraction",
  "library_name": ""
}

{
  "_id": "69e1663e918fd5950ec7440d",
  "id": "cstr/bge-m3-GGUF",
  "modelId": "cstr/bge-m3-GGUF",
  "sha": "18790c04479366de78b255401433434899826c69",
  "createdAt": "2026-04-16T22:44:14.000Z",
  "lastModified": "2026-04-17T07:19:29.000Z",
  "author": "cstr",
  "downloads": 185,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "feature-extraction",
  "library_name": "",
  "siblings_count": 5
}