abiray/qwen3.5-4b-abliterated-claude-4.6-opus-reasoning-distilled-gguf 2 GGUF - Free GGUF Download is indexed on GraySoft with repository links, GGUF quant files, and Hugging Face metadata. This page helps you pick a local model for guIDE or other runtimes. See related models in the same shard below.
Model Intelligence Sheet
abiray/qwen3.5-4b-abliterated-claude-4.6-opus-reasoning-distilled-gguf overview
This is a specialized variant of the Qwen-4B-Reasoning architecture. It has been mathematically modified to neutralize the refusal behaviors and safety guardrails typically found in Claude-distilled reasoning models.
Downloads
414
Likes
0
Pipeline
—
Library
—
Visibility
Public
Access
Open
Repository Files & Downloads
23 files detected
Direct downloads for all repository files
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled.Q4_K_M.gguf | GGUF | Q4_K_M | 2.52 GB | Download |
| Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled.Q5_K_M.gguf | GGUF | Q5_K_M | 2.90 GB | Download |
| Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled.Q6_K.gguf | GGUF | Q6_K | 3.23 GB | Download |
| Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled.Q8_0.gguf | GGUF | — | 4.17 GB | Download |
| Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled.f16.gguf | GGUF | F16 | 7.85 GB | Download |
| ggml-vocab-aquila.gguf | GGUF | — | 4.60 MB | Download |
| ggml-vocab-baichuan.gguf | GGUF | — | 1.28 MB | Download |
| ggml-vocab-bert-bge.gguf | GGUF | — | 0.60 MB | Download |
| ggml-vocab-command-r.gguf | GGUF | — | 10.37 MB | Download |
| ggml-vocab-deepseek-coder.gguf | GGUF | — | 1.10 MB | Download |
| ggml-vocab-deepseek-llm.gguf | GGUF | — | 3.79 MB | Download |
| ggml-vocab-falcon.gguf | GGUF | — | 2.18 MB | Download |
| ggml-vocab-gpt-2.gguf | GGUF | — | 1.68 MB | Download |
| ggml-vocab-gpt-neox.gguf | GGUF | — | 1.69 MB | Download |
| ggml-vocab-llama-bpe.gguf | GGUF | — | 7.46 MB | Download |
| ggml-vocab-llama-spm.gguf | GGUF | — | 0.69 MB | Download |
| ggml-vocab-mpt.gguf | GGUF | — | 1.69 MB | Download |
| ggml-vocab-nomic-bert-moe.gguf | GGUF | — | 6.51 MB | Download |
| ggml-vocab-phi-3.gguf | GGUF | — | 0.69 MB | Download |
| ggml-vocab-qwen2.gguf | GGUF | — | 5.65 MB | Download |
| ggml-vocab-refact.gguf | GGUF | — | 1.64 MB | Download |
| ggml-vocab-starcoder.gguf | GGUF | — | 1.64 MB | Download |
| mmproj-f16.gguf | GGUF | F16 | 641.27 MB | Download |
Model Details Live
Metadata Inspector
Normalized metadata (stored in metadata_json)
{
"metadata": {},
"card_data": {
"license": "other",
"base_model": "Abhiray/Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled",
"tags": [
"abliterated",
"de-censored",
"reasoning",
"qwen",
"distilled",
"deep-scrub"
],
"model_name": "Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled",
"frontmatter": {
"license": "other",
"base_model": "Abhiray/Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled",
"tags": [
"abliterated",
"de-censored",
"reasoning",
"qwen",
"distilled",
"deep-scrub"
],
"model_name": "Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled"
},
"hero_image_url": "",
"summary": "This is a specialized variant of the Qwen-4B-Reasoning architecture. It has been mathematically modified to neutralize the refusal behaviors and safety guardrails typically found in Claude-distilled reasoning models.",
"quick_links": [],
"benchmark_table_html": "",
"readme_markdown": "---\nlicense: other\nbase_model: Abhiray/Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled\ntags:\n- abliterated\n- de-censored\n- reasoning\n- qwen\n- distilled\n- deep-scrub\nmodel_name: Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled\n---\n\n# Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled\n\nThis is a specialized variant of the Qwen-4B-Reasoning architecture. It has been mathematically modified to neutralize the refusal behaviors and safety guardrails typically found in Claude-distilled reasoning models.\n\n## 🛠 The \"Deep-Scrub\" Methodology\n\nStandard abliteration often fails on reasoning models because the \"safety tripwire\" is woven into the early logic chain. This model uses an aggressive early-intercept strategy.\n\n\n### Technical Configuration\n* **Direction Multiplier:** `3.50` (Ultra-Aggressive)\n* **Intervention Range:** `0.05 - 0.95` (Intercepting refusal logic at Layer 2)\n* **Dynamic Layer Targeting:** Enabled (Per-layer refusal vectors)\n* **Hybrid Strategy:** Auto-balanced (Full Attention: 1.0x | Linear Attention: 0.4x)\n* **Refinement:** Winsorization at 0.995 percentile with 0.90 Rank Ratio Null Space Constraints.\n\n\n## 🚀 Key Improvements\n\n1. **Safety Neutralization:** By forcing a **0.05 intercept**, we've targeted the refusal initialization before the model's internal \"Chain of Thought\" can lock onto a refusal state.\n2. **Uninhibited Reasoning:** Designed to bypass the \"However...\" and \"I cannot...\" loops prevalent in distilled reasoning models.\n3. **Architectural Stability:** Despite the high multiplier, we utilized **Norm Preservation** and **Null Space Constraints** to maintain coherence in the model's knowledge base.\n\n\n## ⚠️ Stability & Usage Note\nAt a **3.5x multiplier**, this model is at the upper mathematical limit of stability. \n\n* **Logic Loops:** If you experience \"brain bleed\" (repetitive text), lower your temperature to `0.5 - 0.7`.\n* **System Prompts:** Use an anchoring system prompt to keep the model's logic grounded.\n* **Vision Tasks:** While this is a Vision-Language architecture, the abliteration focused on the **text reasoning layers**.\n\n## ⚖️ Disclaimer\nThis model is provided \"as-is\" for research and creative purposes. The removal of safety guardrails means the user is entirely responsible for the content generated. Please use ethically and responsibly.",
"related_quantizations": []
},
"tags": [
"gguf",
"abliterated",
"de-censored",
"reasoning",
"qwen",
"distilled",
"deep-scrub",
"base_model:Abiray/Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled",
"base_model:quantized:Abiray/Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled",
"license:other",
"endpoints_compatible",
"region:us",
"conversational"
],
"likes": 0,
"downloads": 414,
"gated": false,
"private": false,
"last_modified": "2026-03-10T05:24:54.000Z",
"created_at": "2026-03-09T20:17:40.000Z",
"pipeline_tag": "",
"library_name": ""
}
Source payload excerpt (from Hugging Face API)
{
"_id": "69af2ae4a1c69e6c5047557b",
"id": "Abiray/Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled-GGUF",
"modelId": "Abiray/Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled-GGUF",
"sha": "b74ac84e1e754b78750847e4a95fff054253d73e",
"createdAt": "2026-03-09T20:17:40.000Z",
"lastModified": "2026-03-10T05:24:54.000Z",
"author": "Abiray",
"downloads": 414,
"likes": 0,
"gated": false,
"private": false,
"pipeline_tag": "",
"library_name": "",
"siblings_count": 25
}