llmfan46/Magistral-Small-2509-ultra-uncensored-heretic-v2
Abliterated vs mistralai/Magistral-Small-2509. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 28 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/131, MLP 27/192, Attention 27/256, other 0/4, lm_head 0/1
3 derivatives checked
- mradermacher/Magistral-Small-2509-ultra-uncensored-heretic-v2-i1-GGUF: Quantization (fewer layers)
- mradermacher/Magistral-Small-2509-ultra-uncensored-heretic-v2-GGUF: Quantization (fewer layers)
- llmfan46/Magistral-Small-2509-ultra-uncensored-heretic-v2-GGUF: Same weights, converted (fewer layers)