Question about MLX or raw safetensors release

#5
by apetersson - opened

Hi HauhauCS,

Thanks for publishing these Qwen3.5 models.
Quick question: do you have any plans to release either MLX variants or the uncensored models in raw Hugging Face format (config.json + tokenizer + safetensors), rather than only GGUF?

I’m especially interested in:

  • Qwen3.5-27B-Uncensored-HauhauCS-Aggressive
  • Qwen3.5-122B-A10B-Uncensored-HauhauCS-Aggressive
    I’m asking because MLX conversion on Apple Silicon usually needs the original HF checkpoint, not GGUF.

Thanks again!

Sign up or log in to comment