Qwen3.6 27B Heretic ARA GGUF GGUF

--- library_name: gguf license: apache-2.0 pipeline_tag: text-generation base_model: Qwen/Qwen3.6-27B tags: - GGUF - quantized - heretic - decensored - qwen3.6 - abhiray - mpoa --- # Qwen3.6-27B-heretic (GGUF) This is a decensored version of [Qwen/Qwen3.6-27B](https://huggingface.co/Qwen/Qwen3.6-27B), made using **Heretic v1.2.0** with **Magnitude-Preserving Orthogonal Ablation (MPOA)**. This model is designed for unrestricted, unbound conversational use. ## Abliteration Parameters | Parameter | Value | | :--- | :--- | | `direction_index` | 37.97 | | `attn.o_proj.max_weight` | 1.45 | | `attn.o_proj.max_weight_position` | 59.09 | | `attn.o_proj.min_weight` | 1.44 | | `attn.o_proj.min_weight_distance` | 34.80 | | `mlp.down_proj.max_weight` | 1.43 | | `mlp.down_proj.max_weight_position` | 41.91 | | `mlp.down_proj.min_weight` | 0.72 | | `mlp.down_proj.min_weight_distance` | 28.18 | ## Performance | Metric | This model | Original model (Qwen/Qwen3.6-27B) | | :--- | :--- | :--- | | **KL divergence** | 0.06530 | (by definition) | | **Refusals** | 6/100 | 92/100 | ## Available Quantizations (GGUF) All files are provided in GGUF format, quantized using `llama.cpp` for optimal performance on local hardware. | Filename | Size | | :--- | :--- | | `Qwen3.6-27B-heretic-Q3_K_M.gguf` | 13.3 GB | | `Qwen3.6-27B-heretic-Q4_K_S.gguf` | 15.6 GB | | `Qwen3.6-27B-heretic-Q4_K_M.gguf` | 16.5 GB | | `Qwen3.6-27B-heretic-Q5_K_M.gguf` | 19.2 GB | | `Qwen3.6-27B-heretic-Q6_K.gguf` | 22.1 GB | | `Qwen3.6-27B-heretic-Q8_0.gguf` | 28.6 GB |