ajgazin/Swift-Qwen3.8-27B-Uncensored-MTP
Abliterated vs ukisai/Swift-Qwen3.8-27b. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, o_proj, out_proj) changed, in 64 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Norms 0/325, Attention 65/560, MLP 65/307, Embeddings 1/5, lm_head 0/1, other 0/1
6 derivatives checked
- ajgazin/Swift-Qwen3.8-27B-Uncensored-Dynamic-MTP-GGUF: Same weights, converted
- mradermacher/Swift-Qwen3.8-27B-Uncensored-MTP-GGUF: Quantization
- mradermacher/Swift-Qwen3.8-27B-Uncensored-MTP-i1-GGUF: Quantization
- ajgazin/Swift-Qwen3.8-27B-Uncensored-NVFP4: Quantization
- kylar/Swift-Qwen3.8-27B-Uncensored-MTP-mlx-8Bit: Quantization
- andosen/Swift-Qwen3.8-27B-Uncensored-MTP-mlx-5Bit: Quantization