llmfan46/gemma-4-26B-A4B-it-ultra-uncensored-heretic
Abliterated vs google/gemma-4-26B-A4B-it. Looks like abliteration (refusal removal): only the output projections (o_proj) changed, in 11 of 30 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, MLP 0/231, Norms 0/433, other 0/122, Attention 11/223
10 derivatives checked
- mradermacher/gemma-4-26B-A4B-it-ultra-uncensored-heretic-i1-GGUF: Quantization
- llmfan46/gemma-4-26B-A4B-it-ultra-uncensored-heretic-GGUF: Same weights, converted
- mradermacher/gemma-4-26B-A4B-it-ultra-uncensored-heretic-GGUF: Quantization
- Riyan200324200324/gemma-4-26B-A4B-uncensored-GGUF: Quantization
- akpsahan/gemma-4-26B-A4B-it-ultra-uncensored-heretic-i1-GGUF: Quantization
- louismuk/gemma-4-26B-A4B-heretic-AWQ: Quantization
- JMingo/gemma-4-26B-A4B-it-ultra-uncensored-heretic-UD-GGUF: Same weights, converted
- Arue-2026/gemma-4-26B-A4B-it-ultra-uncensored-heretic-GGUF: Same weights, converted
- culturerevolt/gemma-4-26B-A4B-it-ultra-uncensored-heretic-mlx-4Bit: Quantization
- louismuk/gemma-4-26B-A4B-heretic-NVFP4: Quantization
1 other repos hold byte-for-byte identical weight files.