khtsly/qwen3.5-27b-abliterated-claude-4.6-opus-distilled-32k-gguf overview

{
  "metadata": {},
  "card_data": {
    "language": [
      "en",
      "zh"
    ],
    "license": "apache-2.0",
    "base_model": [
      "huihui-ai/Huihui-Qwen3.5-27B-abliterated"
    ],
    "tags": [
      "unsloth",
      "qwen",
      "qwen3.5",
      "reasoning",
      "chain-of-thought",
      "lora",
      "qlora",
      "luau",
      "abliterated",
      "heretic"
    ],
    "datasets": [
      "nohurry/Opus-4.6-Reasoning-3000x-filtered"
    ],
    "pipeline_tag": "image-text-to-text",
    "frontmatter": {
      "language": [
        "en",
        "zh"
      ],
      "license": "apache-2.0",
      "base_model": [
        "huihui-ai/Huihui-Qwen3.5-27B-abliterated"
      ],
      "tags": [
        "unsloth",
        "qwen",
        "qwen3.5",
        "reasoning",
        "chain-of-thought",
        "lora",
        "qlora",
        "luau",
        "abliterated",
        "heretic"
      ],
      "datasets": [
        "nohurry/Opus-4.6-Reasoning-3000x-filtered"
      ],
      "pipeline_tag": "image-text-to-text"
    },
    "hero_image_url": "",
    "summary": "",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nlanguage:\n- en\n- zh\nlicense: apache-2.0\nbase_model:\n- huihui-ai/Huihui-Qwen3.5-27B-abliterated\ntags:\n- unsloth\n- qwen\n- qwen3.5\n- reasoning\n- chain-of-thought\n- lora\n- qlora\n- luau\n- abliterated\n- heretic\ndatasets:\n- nohurry/Opus-4.6-Reasoning-3000x-filtered\npipeline_tag: image-text-to-text\n---\n\n# Qwen3.5-27B-Abliterated-Claude-4.6-Opus-Distilled-32k\n\n## Performance\n\n| Metric | This model | [Baseline](https://huggingface.co/khtsly/Qwen3.5-27B-Claude-4.6-Opus-Distilled-32k) |\n| :----- | :--------: | :---------------------------: |\n| **KL divergence** | 0.0556 | - |\n| **Refusals** | 12/100 | 94/100 |\n\n## # Training & Abliteration Pipeline Overview\n\n```text\nBase Model (Huihui-Qwen3.5-27B-abliterated)\n │\n ▼\nSupervised Fine-Tuning (SFT) + QLoRA (r=128, α=128, rsLoRA)\n(Response-Only Training masked on \"<|im_start|>assistant\\n\")\n(Max 32k Output Length)\n+\nnohurry/Opus-4.6-Reasoning-3000x-filtered + luau coding samples\n(shuffled)\n │\n ▼\nHeretic\n │\n ▼\nFinal Model (Qwen3.5-27B-Abliterated-Claude-4.6-Opus-Distilled-32k)\n```\n\n### # Training + Filtering Compute & Loss Curve\n* **Hardware:** 1x NVIDIA RTX PRO 6000 Blackwell (96GB)\n* **Total Duration:** ~3-4 Hour\n* **Estimated Total Cost:** $4.00\n* **Distillation Efficacy:** The loss curve demonstrated a strong, healthy downward trajectory throughout the run, confirming successful knowledge transfer from the Opus teacher model. The model converged steadily from an initial loss of **0.579983** down to a final loss of **0.198261**. The final eval loss increases to **~0.22** after **Heretic** filtering phase.\n\n-https://ko-fi.com/khtsly",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "unsloth",
    "qwen",
    "qwen3.5",
    "reasoning",
    "chain-of-thought",
    "lora",
    "qlora",
    "luau",
    "abliterated",
    "heretic",
    "image-text-to-text",
    "en",
    "zh",
    "dataset:nohurry/Opus-4.6-Reasoning-3000x-filtered",
    "base_model:huihui-ai/Huihui-Qwen3.5-27B-abliterated",
    "base_model:adapter:huihui-ai/Huihui-Qwen3.5-27B-abliterated",
    "license:apache-2.0",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 3,
  "downloads": 904,
  "gated": false,
  "private": false,
  "last_modified": "2026-03-08T17:51:14.000Z",
  "created_at": "2026-03-08T15:59:11.000Z",
  "pipeline_tag": "image-text-to-text",
  "library_name": ""
}

{
  "_id": "69ad9ccfb25239d5c9c64f7f",
  "id": "khtsly/Qwen3.5-27B-Abliterated-Claude-4.6-Opus-Distilled-32k-GGUF",
  "modelId": "khtsly/Qwen3.5-27B-Abliterated-Claude-4.6-Opus-Distilled-32k-GGUF",
  "sha": "e49cab1cfd93aee1d5714b586b00edea329bf305",
  "createdAt": "2026-03-08T15:59:11.000Z",
  "lastModified": "2026-03-08T17:51:14.000Z",
  "author": "khtsly",
  "downloads": 904,
  "likes": 3,
  "gated": false,
  "private": false,
  "pipeline_tag": "image-text-to-text",
  "library_name": "",
  "siblings_count": 14
}