TrevorJS/gemma-4-26B-A4B-it-uncensored
Abliterated vs google/gemma-4-26B-A4B-it. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 30 of 30 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, MLP 60/231, Norms 0/433, other 0/122, Attention 30/223
15 derivatives checked
- TrevorJS/gemma-4-26B-A4B-it-uncensored-GGUF: Quantization
- AEON-7/Gemma-4-26B-A4B-it-Uncensored-NVFP4: Quantization
- AEON-7/Gemma-4-E4B-it-Uncensored-NVFP4: Quantization
- BugTraceAI/BugTraceAI-Apex-G4-26B-Q4: Quantization
- BugTraceAI/BugTraceAI-Apex-G4-26B-Master-f16: LoRA-style
- Furude0Rika/Gemma-4-26B-A4B-it-Uncensored-Apex-Quality: Quantization
- manhellis/gemma-4-26B-A4B-it-uncensored-mlx-4Bit: Quantization
- Lookoff/gemma-4-26b-a4b-it-uncensored-4bit: Quantization
- wfiedler/gemma-4-26B-uncensored-mlx-q8: Quantization
- doromiert/gemma-4-26B-A4B-it-uncensored-IQ3_XXS: Quantization
- AgentAnon/gemma-4-26B-A4B-it-uncensored-GGUF: Quantization
- flitsken/herl-gemma: Quantization
- SnowSwordScholar/gemma-4-26B-A4B-it-uncensored-IQ2_M-GGUF: Quantization
- electroglyph/gemma4-26b-fiction-bf16: LoRA-style
- Nabhos/Lehja-RP-Gemma-4-26B-A4B-v0.1: Same weights, repacked
8 other repos hold byte-for-byte identical weight files.