huihui-ai/Huihui-Qwen3.8-27B-abliterated
Abliterated vs Qwen/Qwen3.8-27B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 35 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Norms 0/325, Attention 35/560, MLP 35/307, Embeddings 0/5, lm_head 0/1, other 0/1
74 derivatives checked
- LessThanThreeAI/Qwen3.8-27B-Humanlike-Chat-GGUF: Not comparable yet
- RentedNoodle/Qwen3.8-27B-GSQ-RCO-IQ3_XXS-Uncensored: Quantization
- sakamakismile/Huihui-Qwen3.8-27B-abliterated-NVFP4: Quantization
- TelperionAI/Huihui-Qwen3.8-27B-abliterated-INT4-AWQ-GPTQ: Quantization
- mradermacher/Huihui-Qwen3.8-27B-abliterated-GGUF: Quantization
- prithivMLmods/Qwen3.8-27B-abliterated-GGUF: Abliterated
- ArtomYuan/Qwen3.8-27B-abliterated-ROCmFPX: Same weights, converted
- mradermacher/Huihui-Qwen3.8-27B-abliterated-i1-GGUF: Quantization
- munekazu/Huihui-Qwen3.8-27B-abliterated-FP8: Quantization
- edp1096/Huihui-RadixArk-Qwen3.8-27B-abliterated-NVFP4: Quantization
- lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL: Quantization
- leoncca/Qwen3.8-27B-Huihui-AWQ: Quantization
- ajaeger74/Huihui-Qwen3.8-27B-abliterated-Q6_K-GGUF: Quantization
- AutisticAF/Huihui-Qwen3.8-27B-abliterated-mlx-4Bit: Quantization
- batsclamp/Huihui-Qwen3.8-27B-abliterated-FP8-v2: Quantization
- renketong/Huihui-Qwen3.8-27B-abliterated-NVFP4-GGUF: Quantization
- IsValorum/Huihui-Qwen3.8-27B-Abliterated-VAL-APEX-I-MiniPlus-V2.1-GGUF: Quantization
- Guile/Qwen3.8-27B-Humanlike-Chat-GGUF: LoRA-style
- IsValorum/Huihui-Qwen3.8-27B-Abliterated-VAL-APEX-I-NanoPlus-GGUF: Quantization
- gearwave00001/Huihui-Qwen3.8-27B-Quark-MXFP4-MagicQuant-GGUF: Partial fine-tune
- KostkaIT/Qwen3.8-27B-Huihui-Abliterated-oQ6e-MTP-MLX: Quantization
- mitkox/Huihui-Qwen3.8-27B-abliterated-Q4_K_M-GGUF: Quantization
- lued/Qwen3.8-27B-huihui-abliterated-INT8-W8A16-MTP: Quantization
- augustine223/Huihui-Qwen3.8-27B-abliterated-KO-i1-GGUF: Quantization
- KostkaIT/Qwen3.8-27B-Huihui-Abliterated-oQ4e-MTP-MLX: Quantization
- 0xSojalSec/Qwen3.8-27B-Humanlike-Chat-GGUF: LoRA-style
- prithivMLmods/Qwen3.8-27B-abliterated-MLX: Partial fine-tune
- LessThanThreeAI/Qwen3.8-27B-Humanlike-Chat-2.0: LoRA-style
- ailexleon/Huihui-Qwen3.8-27B-abliterated-mlx-4Bit: Quantization
- Tsotchke-Corporation/Qwen3.8-27B-GeoRefine-D14-abl-GGUF: Quantization
8 other repos hold byte-for-byte identical weight files.