junafinity/Qwen-3.8-27B-Uncensored
Abliterated vs Qwen/Qwen3.8-27B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 48 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Norms 0/325, Attention 48/560, MLP 48/307, Embeddings 0/5, lm_head 0/1, other 0/1
4 derivatives checked
- mradermacher/Qwen-3.8-27B-Uncensored-GGUF: Quantization
- mradermacher/Qwen-3.8-27B-Uncensored-i1-GGUF: Quantization
- junafinity/Qwen-3.8-27B-Uncensored-8-Bit-MLX: Quantization
- tejones36/Qwen-3.8-27B-Uncensored-mlx-mixed_3_6: Quantization
1 other repos hold byte-for-byte identical weight files.