This model has been processed to remove inherent safety filters and refusal mechanisms. It is highly compliant and will generate responses to complex, edge-case, or typically restricted prompts directly from its base weights. No specialized system prompts or catalyst pre-fills are required to bypass refusals.

Available Quantizations

File Name	Format	Description
`qwen3.6-27b-abliterated-Q3_K_M.gguf`	Q3_K_M	Smallest footprint, high perplexity loss. Best for severely RAM-constrained environments.
`qwen3.6-27b-abliterated-Q4_K_S.gguf`	Q4_K_S	Fast inference, slightly lower quality than Q4_K_M.
`qwen3.6-27b-abliterated-Q4_K_M.gguf`	Q4_K_M	Recommended. Excellent balance of performance, size, and quality. (Also available in a dedicated repo).
`qwen3.6-27b-abliterated-Q5_K_M.gguf`	Q5_K_M	High quality, minimal degradation from FP16.
`qwen3.6-27b-abliterated-Q6_K.gguf`	Q6_K	Near-perfect fidelity to original FP16 base. Requires significant RAM.
`qwen3.6-27b-abliterated-Q8_0.gguf`	Q8_0	Maximum quality integer quantization. Very large file size.

Usage with llama.cpp

You can run this model via the command line using standard llama-cli commands. Since the model is abliterated, you do not need to wrap prompts in heavy system instructions.

bash

# Basic inference
./llama-cli -m qwen3.6-27b-abliterated-Q4_K_M.gguf -p "Your prompt here" -n 512 -c 4096

# Interactive conversation mode
./llama-cli -m qwen3.6-27b-abliterated-Q4_K_M.gguf -i -cnv -c 8192

Abiray/Huihui-Qwen3.6-27B-abliterated-GGUF

Qwen 3.6 27B Abliterated - GGUF

Model Details

Abliteration Notes

Available Quantizations

Usage with llama.cpp

Abiray/
Huihui-Qwen3.6-27B-abliterated-GGUF