abiray/huihui-ministral-3b-instruct-2512-abliterated-gguf overview

{
  "metadata": {},
  "card_data": {
    "base_model": "huihui-ai/Huihui-Ministral-3-3B-Instruct-2512-abliterated",
    "library_name": "gguf",
    "license": "apache-2.0",
    "language": [
      "en"
    ],
    "pipeline_tag": "text-generation",
    "tags": [
      "ministral",
      "mistral",
      "quant",
      "gguf",
      "uncensored",
      "abliterated",
      "text-generation-inference",
      "edge-computing"
    ],
    "frontmatter": {
      "base_model": "huihui-ai/Huihui-Ministral-3-3B-Instruct-2512-abliterated",
      "library_name": "gguf",
      "license": "apache-2.0",
      "language": [
        "en"
      ],
      "pipeline_tag": "text-generation",
      "tags": [
        "ministral",
        "mistral",
        "quant",
        "gguf",
        "uncensored",
        "abliterated",
        "text-generation-inference",
        "edge-computing"
      ]
    },
    "hero_image_url": "",
    "summary": "This repository contains **GGUF Quantizations** of the Huihui-Ministral-3B-Instruct-2512-abliterated model. The model is based on **Ministral 3B**, which has been **\"abliterated\"** (uncensored) to remove refusal mechanisms. This modification makes it highly responsive and capable of handling complex, unrestricted creative writing tasks. These GGUF files are strictly optimized for **high-performance local inference** on edge devices, laptops, and consumer-grade hardware.",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nbase_model: huihui-ai/Huihui-Ministral-3-3B-Instruct-2512-abliterated\nlibrary_name: gguf\nlicense: apache-2.0\nlanguage:\n- en\npipeline_tag: text-generation\ntags:\n- ministral\n- mistral\n- quant\n- gguf\n- uncensored\n- abliterated\n- text-generation-inference\n- edge-computing\n---\n\n# Model Card for Huihui-Ministral-3B-Instruct-2512-abliterated-GGUF\n\nThis repository contains **GGUF Quantizations** of the [Huihui-Ministral-3B-Instruct-2512-abliterated](https://huggingface.co/huihui-ai/Huihui-Ministral-3-3B-Instruct-2512-abliterated) model.\n\nThe model is based on **Ministral 3B**, which has been **\"abliterated\"** (uncensored) to remove refusal mechanisms. This modification makes it highly responsive and capable of handling complex, unrestricted creative writing tasks. These GGUF files are strictly optimized for **high-performance local inference** on edge devices, laptops, and consumer-grade hardware.\n\n## Model Details\n\n### Model Description\n\n- **Developed by:** [Abhiray](https://huggingface.co/Abhiray) (Quantization), [Huihui-AI](https://huggingface.co/huihui-ai) (Abliteration), [Mistral AI](https://mistral.ai/) (Base Model)\n- **Model type:** GGUF Quantized LLM (Text Generation)\n- **Language(s) (NLP):** English (Primary), Multilingual capabilities inherited from Ministral.\n- **License:** Apache 2.0\n- **Finetuned from model:** [huihui-ai/Huihui-Ministral-3-3B-Instruct-2512-abliterated](https://huggingface.co/huihui-ai/Huihui-Ministral-3-3B-Instruct-2512-abliterated)\n\n### Model Sources\n\n- **Base Model Repository:** [huihui-ai/Huihui-Ministral-3-3B-Instruct-2512-abliterated](https://huggingface.co/huihui-ai/Huihui-Ministral-3-3B-Instruct-2512-abliterated)\n- **Original Model:** [Mistral AI Ministral-3B](https://mistral.ai/news/ministral-3b/)\n\n## Uses\n\n### Direct Use\n\nThis model is engineered for **efficient local inference** on hardware with limited VRAM. It is compatible with major GGUF inference engines including:\n* **Ollama**\n* **llama.cpp**\n* **LM Studio**\n* **KoboldCPP**\n\nIt is particularly effective for **Creative Writing**, **Interactive Assistants**, and **Narrative Generation** on edge devices where cloud latency or privacy is a concern. The \"abliterated\" nature ensures the model follows instructions precisely without unnecessary refusals.\n\n### Out-of-Scope Use\n\n* **Vision/Image Analysis:** This is a text-only model. It cannot see images.\n* **Fact-Checking:** As a 3B parameter model, it is optimized for creativity and reasoning rather than encyclopedic knowledge retrieval.\n\n## Bias, Risks, and Limitations\n\n**Warning: Uncensored Model**\nThis model has undergone \"abliteration,\" a technique that selectively removes safety guardrails.\n* It **will not refuse** requests that standard models might reject.\n* It may generate sensitive or controversial content if prompted to do so.\n* Users are responsible for the content generated.\n\n## Recommended Stop Tokens\n\nTo prevent the model from generating artifacts (like `+++++`) or hallucinating user replies at the end of a response, ensure your inference tool uses the following stop sequences:\n\n* `</s>`\n* `User:`\n* `Assistant:`\n\n## How to Get Started with the Model\n\n### Option 1: Run with Ollama (Easiest)\nYou can pull this model directly to your command line:\n\n```bash\nollama run hf.co/Abhiray/Huihui-Ministral-3B-Instruct-2512-abliterated-GGUF:Q4_K_M",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "ministral",
    "mistral",
    "quant",
    "uncensored",
    "abliterated",
    "text-generation-inference",
    "edge-computing",
    "text-generation",
    "en",
    "base_model:huihui-ai/Huihui-Ministral-3-3B-Instruct-2512-abliterated",
    "base_model:quantized:huihui-ai/Huihui-Ministral-3-3B-Instruct-2512-abliterated",
    "license:apache-2.0",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 3,
  "downloads": 246,
  "gated": false,
  "private": false,
  "last_modified": "2026-01-20T07:09:37.000Z",
  "created_at": "2026-01-20T04:45:26.000Z",
  "pipeline_tag": "text-generation",
  "library_name": "gguf"
}

{
  "_id": "696f0866efd818e4a9b9baf2",
  "id": "Abiray/Huihui-Ministral-3B-Instruct-2512-abliterated-GGUF",
  "modelId": "Abiray/Huihui-Ministral-3B-Instruct-2512-abliterated-GGUF",
  "sha": "aaa02c2d471c4702ea95b2c59c8c00beda4dde95",
  "createdAt": "2026-01-20T04:45:26.000Z",
  "lastModified": "2026-01-20T07:09:37.000Z",
  "author": "Abiray",
  "downloads": 246,
  "likes": 3,
  "gated": false,
  "private": false,
  "pipeline_tag": "text-generation",
  "library_name": "gguf",
  "siblings_count": 7
}