Why this repository?
Unlike incomplete GGUF uploads, this repository provides the full 11-quantization spectrum (from high-precision F16 down to lightweight Q2_K) of the abliterated Qwen3-Coder-30B-A3B-Instruct model.
Choose the exact fit for your VRAM/RAM constraints without sacrificing reasoning capabilities.
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated - GGUF
This is the integral and complete collection of quantizations in GGUF format for the Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated model, prepared locally for use with llama.cpp, Ollama, LM Studio, or Text-Generation-WebUI.
This model combines the advanced programming and logical reasoning capabilities of the Qwen3-Coder (MoE) architecture, but processed with abliteration techniques to root out artificial system censorship filters and blocks, responding without restrictions.
📋 Available Files (Complete Collection Without Splits)
| File | Est. Size | BPW (Bits per Weight) | Recommended Usage Profile |
|---|---|---|---|
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-F16.gguf |
~61.1 GB | 16.00 | Complete Base Template. Absolute fidelity of floating-point weights. |
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-Q8_0.gguf |
~32.5 GB | 8.50 | Identical quality to the original, ideal for maximizing performance in advanced local environments with high-end GPUs. |
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-Q6_K.gguf |
~25.1 GB | 6.59 | Extremely high retention of programming syntax and logic with an optimized weight. |
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-Q5_K_M.gguf |
~21.7 GB | 5.69 | Recommended Sweet Spot. Keeps code coherence intact while critically reducing weight. |
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-Q5_K_S.gguf |
~21.1 GB | 5.54 | Compact 5-bit variant. |
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-Q4_K_M.gguf |
~18.6 GB | 4.85 | The Most Wanted. Optimal balance for running software development inferences on advanced home setups. |
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-Q4_K_S.gguf |
~17.5 GB | 4.58 | Compact 4-bit variant to accelerate tokens-per-second speed. |
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-Q3_K_L.gguf |
~15.9 GB | 4.01 | Medium-high 3-bit compression. Retains basic programming logic. |
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-Q3_K_M.gguf |
~14.7 GB | 3.66 | Intermediate 3-bit variant. |
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-Q3_K_S.gguf |
~13.3 GB | 3.44 | Lightweight 3-bit variant. |
Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-Q2_K.gguf |
~11.3 GB | 2.90 | Extreme Compression. May experience structure degradation and complex code bleeding. For experimental development only. |
Note: Sizes are initial baseline estimates based on the model's native weight in safetensors (31B MoE parameters); verifying the final size on disk after local compilation is recommended.
💡 Highlighted Usage
Since this model is optimized for coding tasks and uncensored instructions, you can call it directly in your terminal using llama.cpp:
./llama-cli -m Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-Q4_K_M.gguf -n 2048 -p "Write a Python script to perform advanced web scraping while avoiding blocks."
If you prefer using Ollama, you can run it directly with:
ollama run huihui_ai/qwen3-coder-abliterated
⚖️ Disclaimer
This model is open-access and lacks artificial safety filters (abliterated). Its original alignment has been modified for experimental and research purposes. Generated content, whether text responses or software code, is the sole responsibility of the individual running the local inference.