CoolFace
Modelpublic

cesarsal1nas/Huihui-Qwen3.5-35B-A3B-abliterated-Q4_K_M-GGUF

sourceHugging Faceapache-2.0updated 7mo agoView on Hugging Face
2likes315downloads
Model Card
[!WARNING] A significantly improved version of this model is available. This repo quantizes huihui-ai/Huihui-Qwen3.5-35B-A3B-abliterated — the standard abliteration. A newer fine-tune of the same architecture, trained in the style of Claude 4.6 Opus, has since been released and produces noticeably richer, more expressive outputs. ➡️ Recommended upgrade: [Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-Q4_K_M-GGUF](https://huggingface.co/cesarsal1nas/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-Q4_K_M-GGUF) Same architecture, same Q4KM quantization, same VRAM footprint — just a better fine-tune. This repo will remain available for reference.

Huihui-Qwen3.5-35B-A3B-abliterated — Q4KM GGUF

This is a Q4KM GGUF quantization of huihui-ai/Huihui-Qwen3.5-35B-A3B-abliterated.

Refer to the original model card for full details, usage warnings, and licensing information.

Details

PropertyValue
Source modelhuihui-ai/Huihui-Qwen3.5-35B-A3B-abliterated
Architectureqwen35moe (35B total params, ~3B active; 256 experts, 8 active per token)
QuantizationQ4KM (~4.8 BPW)
File size~21 GB
Quantized withllama.cpp

Usage with llama.cpp

bash
llama-cli \
  --hf-repo cesarsal1nas/Huihui-Qwen3.5-35B-A3B-abliterated-Q4_K_M-GGUF \
  --hf-file huihui-qwen3.5-35b-a3b-abliterated-Q4_K_M.gguf \
  -p "Tell me about the universe"
bash
llama-server \
  --hf-repo cesarsal1nas/Huihui-Qwen3.5-35B-A3B-abliterated-Q4_K_M-GGUF \
  --hf-file huihui-qwen3.5-35b-a3b-abliterated-Q4_K_M.gguf \
  -c 8192

Usage with Ollama

Requires Ollama with qwen35moe support. See PR #14506 for the architecture patch.

bash
ollama run hf.co/cesarsal1nas/Huihui-Qwen3.5-35B-A3B-abliterated-Q4_K_M-GGUF

Credits