lhca521/minimax-m2.7-abliterated-heretic-gguf overview

{
  "metadata": {},
  "card_data": {
    "base_model": "MiniMaxAI/MiniMax-M2.7",
    "library_name": "gguf",
    "pipeline_tag": "text-generation",
    "license": "other",
    "license_name": "non-commercial",
    "license_link": "https://github.com/MiniMax-AI/MiniMax-M2.7/blob/main/LICENSE",
    "tags": [
      "gguf",
      "minimax",
      "minimax_m2",
      "moe",
      "mixture-of-experts",
      "abliterated",
      "uncensored",
      "heretic",
      "ara",
      "llama-cpp"
    ],
    "quantized_by": "Youssofal",
    "frontmatter": {
      "base_model": "MiniMaxAI/MiniMax-M2.7",
      "library_name": "gguf",
      "pipeline_tag": "text-generation",
      "license": "other",
      "license_name": "non-commercial",
      "license_link": "https://github.com/MiniMax-AI/MiniMax-M2.7/blob/main/LICENSE",
      "tags": [
        "gguf",
        "minimax",
        "minimax_m2",
        "moe",
        "mixture-of-experts",
        "abliterated",
        "uncensored",
        "heretic",
        "ara",
        "llama-cpp"
      ],
      "quantized_by": "Youssofal"
    },
    "hero_image_url": "",
    "summary": "This is a GGUF release of an abliterated version of MiniMaxAI's MiniMax-M2.7. By applying Heretic's Ablated Refusal Adaptation (ARA), the base refusal behavior was removed at the weight level. The result keeps MiniMax-M2.7's sparse MoE reasoning, long-context instruction following, and general capability profile, but no longer defaults to the original refusal pattern.",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nbase_model: MiniMaxAI/MiniMax-M2.7\nlibrary_name: gguf\npipeline_tag: text-generation\nlicense: other\nlicense_name: non-commercial\nlicense_link: https://github.com/MiniMax-AI/MiniMax-M2.7/blob/main/LICENSE\ntags:\n  - gguf\n  - minimax\n  - minimax_m2\n  - moe\n  - mixture-of-experts\n  - abliterated\n  - uncensored\n  - heretic\n  - ara\n  - llama-cpp\nquantized_by: Youssofal\n---\n\n# MiniMax-M2.7-Abliterated-Heretic-GGUF\n\nThis is a GGUF release of an abliterated version of MiniMaxAI's MiniMax-M2.7.\n\nBy applying Heretic's Ablated Refusal Adaptation (ARA), the base refusal behavior was removed at the weight level. The result keeps MiniMax-M2.7's sparse MoE reasoning, long-context instruction following, and general capability profile, but no longer defaults to the original refusal pattern.\n\n## Methodology & Model Notes\n\nMiniMax-M2.7 is a 229B sparse MoE model with 10B active parameters per token, 62 layers, hybrid attention, 256 local experts with 8 active per token, and a 200K context window.\n\nThis release was produced with a direct Heretic ARA run using the fixed parameter set below:\n\n- `start_layer_index = 30`\n- `end_layer_index = 51`\n- `preserve_good_behavior_weight = 0.4512`\n- `steer_bad_behavior_weight = 0.0037`\n- `overcorrect_relative_weight = 0.8804`\n- `neighbor_count = 14`\n\nThe direct ARA run completed with `Refusals: 0/25`.\n\nThe resulting abliterated checkpoint was exported to BF16 and then converted to GGUF for llama.cpp-compatible deployment.\n\n## Files\n\n- `MiniMax-M2.7-abliterated-BF16/`: BF16 GGUF split into 10 parts\n- `MiniMax-M2.7-abliterated-Q8_0/`: Q8_0 GGUF split into 5 parts\n- `MiniMax-M2.7-abliterated-Q3_K_M/`: Q3_K_M GGUF split for Hub delivery\n- Additional quants will be added from the same abliterated BF16 GGUF source\n\n## Prompt Format\n\n```text\n]~!b[]~b]system\n{system_prompt}[e~[\n]~b]user\n{prompt}[e~[\n]~b]ai\n<think>\n```\n\n## Running\n\n```bash\nllama-server \\\n  -m <quant-file.gguf> \\\n  -ngl 999 -c 32768 --jinja \\\n  --reasoning-format auto -fa \\\n  --temp 1.0 --top-p 0.95 --top-k 40\n```\n\n## Model Architecture\n\n| Spec | Value |\n|---|---|\n| Total Parameters | 229B (sparse MoE) |\n| Active Parameters | 10B per token |\n| Experts | 256 local, 8 per token |\n| Layers | 62 |\n| Attention | Hybrid: 7 Lightning + 1 softmax per 8-block |\n| Context | 200K |\n| Base Model | MiniMaxAI/MiniMax-M2.7 |\n\n## Disclaimer\n\nThis model has had refusal behavior removed at the weight level. It will answer prompts that the base model would normally refuse. You are responsible for how you use it.\n\n## Credits\n\n- Base model: [MiniMaxAI/MiniMax-M2.7](https://huggingface.co/MiniMaxAI/MiniMax-M2.7)\n- Refusal removal pipeline: [Heretic](https://github.com/andyrdt/heretic) with the ARA method\n- GGUF runtime and quantization: [llama.cpp](https://github.com/ggml-org/llama.cpp)\n\n## License\n\nThis release inherits the base MiniMax-M2.7 license.\n\n**NON-COMMERCIAL.** Commercial use requires written authorization from MiniMax.\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "minimax",
    "minimax_m2",
    "moe",
    "mixture-of-experts",
    "abliterated",
    "uncensored",
    "heretic",
    "ara",
    "llama-cpp",
    "text-generation",
    "base_model:MiniMaxAI/MiniMax-M2.7",
    "base_model:quantized:MiniMaxAI/MiniMax-M2.7",
    "license:other",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 0,
  "downloads": 222,
  "gated": false,
  "private": false,
  "last_modified": "2026-04-15T13:41:50.000Z",
  "created_at": "2026-04-15T13:41:50.000Z",
  "pipeline_tag": "text-generation",
  "library_name": "gguf"
}

{
  "_id": "69df959e0373e07ad9a11065",
  "id": "lhca521/MiniMax-M2.7-Abliterated-Heretic-GGUF",
  "modelId": "lhca521/MiniMax-M2.7-Abliterated-Heretic-GGUF",
  "sha": "7670790baedc41e7e0b8a6e038bc88ba51856e44",
  "createdAt": "2026-04-15T13:41:50.000Z",
  "lastModified": "2026-04-15T13:41:50.000Z",
  "author": "lhca521",
  "downloads": 222,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "text-generation",
  "library_name": "gguf",
  "siblings_count": 31
}