coder3101/gemma-4-12B-it-heretic
Abliterated vs google/gemma-4-12B-it. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 43 of 48 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/12, Norms 0/289, other 0/48, MLP 42/144, Attention 43/184
7 derivatives checked
- mradermacher/gemma-4-12B-it-heretic-i1-GGUF: Quantization
- JMingo/gemma-4-12B-it-heretic-UD-GGUF: Same weights, converted
- mradermacher/gemma-4-12B-it-heretic-GGUF: Quantization
- artokun/gemma4-comfyui-mcp: Not comparable yet
- thesamgodson/Gemma-4-12B-Hephaestus-GGUF: Quantization
- 21world/gemma4-comfyui-mcp: Not comparable yet
- xdavxd/gemma-4-12B-it-heretic-NVFP4: Quantization