llmfan46/Qwen3.6-35B-A3B-uncensored-heretic
Abliterated vs Qwen/Qwen3.6-35B-A3B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 34 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Attention 34/280, MLP 60/280, Norms 0/131, lm_head 0/1
24 derivatives checked
- llmfan46/Qwen3.6-35B-A3B-uncensored-heretic-GGUF: Abliterated
- mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF: Quantization
- llmfan46/Qwen3.6-35B-A3B-uncensored-heretic-NVFP4-Experts-Only-GGUF: Quantization
- mradermacher/Qwen3.6-35B-A3B-uncensored-heretic-i1-GGUF: Full fine-tune
- froggeric/Qwen3.6-35B-A3B-Uncensored-Heretic-MLX-4bit: Quantization
- llmfan46/Qwen3.6-35B-A3B-uncensored-heretic-GPTQ-Int4: Quantization
- peculiar-ragdoll/Occult-Nail-1.0-35B-A3B-GGUF: Quantization
- SassyDiffusion/Qwen3.6-35B-A3B-heretic-GGUF: Same weights, converted
- froggeric/Qwen3.6-35B-A3B-Uncensored-Heretic-MLX-8bit: Quantization
- protoLabsAI/Qwen3.6-35B-A3B-uncensored-heretic-FP8: Quantization
- llmfan46/Qwen3.6-35B-A3B-uncensored-heretic-NVFP4-Experts-Only: Quantization
- noctrex/Qwen3.6-35B-A3B-uncensored-heretic-MXFP4_MOE-GGUF: Abliterated
- sahilchachra/Qwen3.6-35B-A3B-uncensored-heretic-AWQ: Quantization
- magiccodingman/Qwen3.6-35B-A3B-Uncensored-MagicQuant-GGUF: Quantization
- froggeric/Qwen3.6-35B-A3B-Uncensored-Heretic-MLX-6bit: Quantization
- FoolDev/Janus-35B-HERETIC: Quantization
- k0ndra/Qwen3.6-35B-A3B-uncensored-heretic-ja-imatrix-GGUF: Quantization
- BlueBackup/Qwen3.6-35B-A3B-uncensored-heretic-IQ2_M: Quantization
- marafx2025/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF: Quantization
- PrismLive/PrismLive-35B: Quantization
- claymorecrystal/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF: Quantization
- WeilPesi/Qwen3.6-35B-A3B-uncensored-heretic-NVFP4-Experts-Only-GGUF: Quantization
- TopherAU/Qwen3.6-35B-A3B-uncensored-heretic-GGUF: Quantization
- LeaderboardModel1/Qwen3.6-35B-A3B-uncensored-heretic-AutoRound-W4A16-RTN: Quantization
2 other repos hold byte-for-byte identical weight files.