The world's first reasoning-enhanced uncensored 397B model. Abliterated + LoRA fine-tuned on 12,842 high-quality reasoning samples distilled from Anthropic's Opus 4.6 outputs.
This is Stage 2 of the Qwen3.5-397B pipeline:
397B total parameters, 17B active per token (Mixture-of-Experts). Trained for 3,046 steps across 8×H200 GPUs. Final loss: 0.363, accuracy: 90.2%.
| Stage 1 (Abliterated Only) | Stage 2 (This Model) | |
|---|---|---|
| Abliteration | ✅ Custom pipeline | ✅ Same pipeline |
| Fine-tuning | None | LoRA r=64, 134.7M trainable params |
| Training data | None | 12,842 reasoning samples (Opus 4.6 distillation) |
| Reasoning quality | Base Qwen3.5 | Enhanced chain-of-thought + structured analysis |
| Thinking mode | Default | Trained with <think> tags for explicit reasoning |
| Final loss | N/A | 0.363 |
| Final accuracy | N/A | 90.2% |
LoRA Configuration:
self_attn.{q,k,v,o}_proj, shared_expert.{gate,up,down}_projTraining Hyperparameters:
Dataset Composition (12,842 samples, deduplicated):
opus-10000x: 9,633 multi-turn conversations with deep reasoningopus-3000x: 2,326 problem/thinking/solution samples with explicit chain-of-thoughtreasoning-700x: 633 complex reasoning and analytical taskshigh-reasoning-250x: 250 elite-tier reasoning samples requiring multi-step deductionAll samples feature reasoning traces distilled from Anthropic's Claude Opus 4.6, including <think> tag formatting for explicit chain-of-thought.
Training Curve:
| Quant | Size | BPW | RAM Required | Description | Use Case |
|---|---|---|---|---|---|
| BF16 | 739 GB (2 splits) | 16.01 | ~750 GB | Full precision | Reference, maximum quality |
| Q8_0 | 393 GB | 8.51 | ~400 GB | 8-bit | Best quality with compression |
| Q6_K | 304 GB | 6.57 | ~310 GB | 6-bit | High quality, good compression |
| Q4_K_M | 225 GB | ~4.85 | ~230 GB | 4-bit mixed | Recommended for most users |
| Q3_K_M | 177 GB | ~3.83 | ~185 GB | 3-bit mixed | Memory-constrained setups |
| Q2_K | 135 GB | ~2.92 | ~140 GB | 2-bit | Extreme compression |
Note: Q5_K_M is unavailable due to infrastructure loss during upload. Will be regenerated and uploaded in a future update.
# Recommended: Q4_K_M for balanced quality/memory
./llama-cli -m Qwen3.5-397B-A17B-Opus-4.6-Reasoning-Uncensored-Q4_K_M.gguf \
-p "You are a helpful uncensored assistant with strong reasoning abilities." \
-n 2048 --temp 0.7 --top-p 0.9
# Server mode with large context
./llama-server -m Qwen3.5-397B-A17B-Opus-4.6-Reasoning-Uncensored-Q4_K_M.gguf \
--port 8080 --host 0.0.0.0 -c 131072
Download the GGUF file and load it in LM Studio. The model supports <think> reasoning tags — enable thinking mode for best results on complex tasks.
Point your backend to a llama.cpp server. Full OpenAI-compatible API at /v1/chat/completions.
Qwen3.5-397B-A17B (base)
↓ Custom abliteration (strength 20.0, attn.o_proj + shared_expert.down_proj)
↓ LoRA fine-tuning (12,842 Opus 4.6 reasoning samples, 3,046 steps)
↓ LoRA merge into base weights
↓ BF16 GGUF conversion (llama.cpp)
↓ Quantization cascade (Q8_0 → Q2_K)
All processing done in BF16 on 8×H200 SXM5 GPUs (1.1TB VRAM total). Abliteration and quantization applied in correct order: full-precision abliteration → training → merge → THEN quantize.
<think> tags for reasoning. Strip them in post-processing if unwanted.Every donation helps fund more open-weight model releases. ⚡ Forged on 8×NVIDIA H200 SXM5 | 1.1TB VRAM
| Currency | Address |
|---|---|
| BTC | bc1p4q7vpwucvww2y3x4nhps4y4vekye8uwm9re5a0kx8l6u5nky5ucszm2qhh |
| ETH | 0xe5Aa16E53b141D42458ABeEDb00a157c3Fea2108 |
| SOL | 9CXwjG1mm9uLkxRevdMQiF61cr6TNHSiWtFRHmUEgzkG |
Qwen3.5-397B-A17B-Opus-4.6-Reasoning-Uncensored-BF16-split--00001-of-00002.gguf 399 GBQwen3.5-397B-A17B-Opus-4.6-Reasoning-Uncensored-BF16-split--00002-of-00002.gguf 394 GBHave the files? Submit a torrent.
Submit a torrentHave the complete files and permission to share them? Replace the example folder path. The script checks the files, makes a torrent and submits its magnet and checksums. Keep your torrent client seeding. Get your community key from your settings, and read the script before running it.
curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo timteh673/Qwen3.5-397B-A17B-Opus-4.6-Reasoning-Uncensored-GGUF ./model-folderQwen/Qwen3.5-397B-A17B