llmfan46/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved
Abliterated vs Qwen/Qwen3.6-27B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 63 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/5, MLP 64/307, Attention 56/560, lm_head 0/1, Norms 0/325, other 0/1
13 derivatives checked
- llmfan46/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-GGUF: Same weights, converted
- mradermacher/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-i1-GGUF: Quantization
- magiccodingman/Qwen3.6-27B-Uncensored-MagicQuant-MTP-GGUF: Quantization
- mradermacher/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-GGUF: Quantization
- llmfan46/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-NVFP4-GGUF: Quantization
- Stabhappy/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-GGUF: Quantization
- llmfan46/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-GPTQ-Int4: Quantization
- texdata/Qwen3.6-27B-slo-med-mt-GGUF-Q8: Full fine-tune
- cvgro/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-GGUF: Same weights, converted
- llmfan46/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-NVFP4: Quantization
- Gryphe/Pantheon-Reasoning-27B: Full fine-tune
- underlotus/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-oQ4-mtp: Quantization
- abrasdaosfjnps/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-GGUF: Same weights, converted