llmfan46/gemma-4-31B-it-uncensored-heretic
Abliterated vs google/gemma-4-31B-it. Looks like abliteration (refusal removal): only the output projections (o_proj) changed, in 18 of 60 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/523, other 0/62, MLP 0/261, Attention 18/338
22 derivatives checked
- llmfan46/gemma-4-31B-it-uncensored-heretic-GGUF: Same weights, converted (fewer layers)
- mradermacher/gemma-4-31B-it-uncensored-heretic-i1-GGUF: Quantization (fewer layers)
- llmfan46/gemma-4-31B-it-uncensored-heretic-NVFP4-GGUF: Same weights, converted
- JMingo/gemma-4-31B-it-uncensored-heretic-UD-GGUF: Same weights, converted (fewer layers)
- mlx-community/gemma-4-31B-it-uncensored-heretic-4bit: Quantization
- mradermacher/gemma-4-31B-it-uncensored-heretic-GGUF: Quantization (fewer layers)
- Riyan200324200324/gemma-4-31B-it-uncensored-heretic-GGUF: Same weights, converted (fewer layers)
- okanity/gemma-4-31B-it-uncensored-heretic-GGUF: Same weights, converted (fewer layers)
- valoomba/Gemma-4-31B-it-uncensored-heretic-Quark-W8A8-INT8: Quantization
- llmfan46/gemma-4-Ortenzya-The-Creative-Wordsmith-31B-it-uncensored-heretic: LoRA-style
- felixwll/gemma-4-31B-it-uncensored-heretic-GGUF: Same weights, converted (fewer layers)
- kakrotto/gemma-4-31B-it-uncensored-heretic-FP8: Quantization
- llmfan46/gemma-4-31B-it-uncensored-heretic-NVFP4: Quantization
- lactroiii/gemma-4-31B-it-uncensored-heretic-GGUF: Quantization (fewer layers)
- vwdubb/gemma-4-31B-it-uncensored-heretic-FP8: Quantization
- Andu-bit/gemma-4-31B-it-uncensored-heretic-GGUF: Same weights, converted (fewer layers)
- pupsipups/gemma-4-31B-it-uncensored-heretic-GGUF: Same weights, converted (fewer layers)
- ailexleon/gemma-4-31B-it-uncensored-heretic-mlx-vlm-8Bit: Quantization
- ailexleon/gemma-4-31B-it-uncensored-heretic-mlx-lm-8Bit: Quantization
- Yunshanccx/gemma-4-31B-it-uncensored-heretic-GGUF: Same weights, converted (fewer layers)
- HerrisII/gemma-4-31B-it-uncensored-heretic-Q8_0-GGUF: Quantization (fewer layers)
- worstplayer/gemma-4-31B-it-uncensored-heretic-IQ3_XS.gguf: Quantization (fewer layers)
5 other repos hold byte-for-byte identical weight files.