huihui-ai/Huihui-gemma-4-26B-A4B-it-abliterated
Abliterated vs google/gemma-4-26B-A4B-it. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 25 of 30 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, MLP 25/231, Norms 0/433, other 0/122, Attention 25/223
14 derivatives checked
- groxaxo/Huihui-gemma-4-26B-A4B-it-abliterated-GGUF: Quantization
- mradermacher/Huihui-gemma-4-26B-A4B-it-abliterated-i1-GGUF: Quantization
- mradermacher/Huihui-gemma-4-26B-A4B-it-abliterated-GGUF: Quantization
- deewu0809/gemma-4-26B-A4B-it-abliterated-GGUF: Quantization
- leonsarmiento/Huihui-gemma-4-26B-A4B-it-abliterated-3bit-XL-mlx: Quantization
- vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-4bit: Quantization
- sakamakismile/Huihui-gemma-4-26B-A4B-it-abliterated-NVFP4: Quantization
- coolthor/Huihui-gemma-4-26B-A4B-it-abliterated-FP8-Dynamic: Quantization
- vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-8bit: Quantization
- vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16: Different architecture
- hehua2008/gemma-4-26B-A4B-it-abliterated-MLX-4bit: Quantization
- raver1975/Huihui-gemma-4-26B-A4B-it-abliterated-4bit: Quantization
- leonsarmiento/Huihui-gemma-4-26B-A4B-it-6bit-XL-mlx: Quantization
- win10/K2-nano-preview: Identical copy
6 other repos hold byte-for-byte identical weight files.