richardyoung/qwen2.5-7b-instruct-abliterated-gguf overview

{
  "metadata": {},
  "card_data": {
    "license": "apache-2.0",
    "base_model": "Qwen/Qwen2.5-7B-Instruct",
    "tags": [
      "gguf",
      "abliterated",
      "uncensored",
      "qwen2.5",
      "llama-cpp",
      "ollama",
      "lm-studio",
      "quantized"
    ],
    "pipeline_tag": "text-generation",
    "language": [
      "en"
    ],
    "frontmatter": {
      "license": "apache-2.0",
      "base_model": "Qwen/Qwen2.5-7B-Instruct",
      "tags": [
        "gguf",
        "abliterated",
        "uncensored",
        "qwen2.5",
        "llama-cpp",
        "ollama",
        "lm-studio",
        "quantized"
      ],
      "pipeline_tag": "text-generation",
      "language": [
        "en"
      ]
    },
    "hero_image_url": "",
    "summary": "An **abliterated** (uncensored) version of Qwen/Qwen2.5-7B-Instruct in GGUF format, ready for local inference with llama.cpp, Ollama, or LM Studio. **Abliteration** removes the refusal training from the model while preserving its core capabilities — useful for research, creative writing, and scenarios where you need unrestricted model output.",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nlicense: apache-2.0\nbase_model: Qwen/Qwen2.5-7B-Instruct\ntags:\n  - gguf\n  - abliterated\n  - uncensored\n  - qwen2.5\n  - llama-cpp\n  - ollama\n  - lm-studio\n  - quantized\npipeline_tag: text-generation\nlanguage:\n  - en\n---\n\n# Qwen2.5-7B-Instruct Abliterated (GGUF)\n\nAn **abliterated** (uncensored) version of [Qwen/Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct) in GGUF format, ready for local inference with llama.cpp, Ollama, or LM Studio.\n\n**Abliteration** removes the refusal training from the model while preserving its core capabilities — useful for research, creative writing, and scenarios where you need unrestricted model output.\n\n## Quick Start\n\n### With Ollama\n\n```bash\nollama run hf.co/richardyoung/Qwen2.5-7B-Instruct-abliterated-GGUF\n```\n\n### With llama.cpp\n\n```bash\n# Download the Q4_K_M quantization (recommended balance of quality/speed)\nhuggingface-cli download richardyoung/Qwen2.5-7B-Instruct-abliterated-GGUF \\\n    --include \"*Q4_K_M*\" --local-dir ./models\n\n# Run inference\n./llama-cli -m ./models/*Q4_K_M*.gguf \\\n    -p \"You are a helpful assistant.\" \\\n    --chat-template chatml -ngl 99\n```\n\n### With LM Studio\n\nSearch for `richardyoung/Qwen2.5-7B-Instruct-abliterated-GGUF` in the model browser, or download manually and import.\n\n## Available Quantizations\n\n| Quantization | Use Case |\n|---|---|\n| Q2_K | Minimum RAM, lower quality |\n| Q4_K_M | **Recommended** — good balance of quality and speed |\n| Q5_K_M | Higher quality, more RAM |\n| Q6_K | Near-original quality |\n| Q8_0 | Maximum quality, most RAM |\n\n## What is Abliteration?\n\nAbliteration is a technique that identifies and removes the \"refusal direction\" in a model's residual stream. Unlike fine-tuning, it surgically modifies the model's behavior without retraining, preserving the original model's knowledge and capabilities.\n\nFor more details, see the original research: [Refusal in Language Models Is Mediated by a Single Direction](https://arxiv.org/abs/2406.11717)\n\n## Intended Use\n\nThis model is intended for:\n- **Research** on model alignment and safety\n- **Creative writing** without artificial restrictions\n- **Education** on how language model censorship works\n- **Local inference** where you control the deployment context\n\n## Limitations\n\n- Abliterated models will comply with requests the base model would refuse\n- Use responsibly — the model has no safety guardrails\n- Output quality matches the base Qwen2.5-7B-Instruct\n\n## Other Models by richardyoung\n\n- **Abliterated/Uncensored models**: [Qwen2.5-7B](https://hf.co/richardyoung/Qwen2.5-7B-Instruct-abliterated-GGUF) | [Qwen3-14B](https://hf.co/richardyoung/Qwen3-14B-abliterated-GGUF) | [DeepSeek-R1-32B](https://hf.co/richardyoung/Deepseek-R1-Distill-Qwen-32b-uncensored) | [Qwen3-8B](https://hf.co/richardyoung/Qwen3-8B-Abliterated)\n- **MLX quantizations (Apple Silicon)**: [Kimi-K2 series](https://hf.co/richardyoung/Kimi-K2-Instruct-0905-MLX-4bit) | [olmOCR MLX](https://hf.co/richardyoung/olmOCR-2-7B-1025-MLX-4bit)\n- **OCR & Vision**: [olmOCR GGUF](https://hf.co/richardyoung/olmOCR-2-7B-1025-GGUF)\n- **Healthcare/Medical**: [Synthea 575K patients dataset](https://hf.co/datasets/richardyoung/synthea-575k-patients) | [CardioEmbed](https://hf.co/richardyoung/CardioEmbed)\n- **Research**: [LLM Instruction-Following Evaluation](https://hf.co/richardyoung/llm-instruction-following-paper) (arxiv:2510.18892)\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "abliterated",
    "uncensored",
    "qwen2.5",
    "llama-cpp",
    "ollama",
    "lm-studio",
    "quantized",
    "text-generation",
    "en",
    "arxiv:2406.11717",
    "arxiv:2510.18892",
    "base_model:Qwen/Qwen2.5-7B-Instruct",
    "base_model:quantized:Qwen/Qwen2.5-7B-Instruct",
    "license:apache-2.0",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 0,
  "downloads": 566,
  "gated": false,
  "private": false,
  "last_modified": "2026-03-23T02:06:46.000Z",
  "created_at": "2025-12-01T16:37:08.000Z",
  "pipeline_tag": "text-generation",
  "library_name": ""
}

{
  "_id": "692dc4345ca017776273d36a",
  "id": "richardyoung/Qwen2.5-7B-Instruct-abliterated-GGUF",
  "modelId": "richardyoung/Qwen2.5-7B-Instruct-abliterated-GGUF",
  "sha": "f9e76b06f186e96e85361ece63e34619b7d3217b",
  "createdAt": "2025-12-01T16:37:08.000Z",
  "lastModified": "2026-03-23T02:06:46.000Z",
  "author": "richardyoung",
  "downloads": 566,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "text-generation",
  "library_name": "",
  "siblings_count": 5
}