llmfan46/MiniMax-M2.7-BF16-ultra-uncensored-heretic
Abliterated vs amd/MiniMax-M2.7-BF16. Looks like abliteration (refusal removal): only the output projections (o_proj) changed, in 34 of 62 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: other 0/124, Embeddings 0/1, MLP 0/2377, Norms 0/249, Attention 34/248, lm_head 0/1
4 derivatives checked
- llmfan46/MiniMax-M2.7-ultra-uncensored-heretic-GGUF: Same weights, converted
- mradermacher/MiniMax-M2.7-BF16-ultra-uncensored-heretic-i1-GGUF: Quantization
- dawncr0w/MiniMax-M2.7-ultra-uncensored-heretic-oQ4-MLX: Quantization
- alonsoko/MiniMax-M2.7-BF16-ultra-uncensored-heretic-AWQ-W4A16: Quantization
1 other repos hold byte-for-byte identical weight files.