Declare compatibility with the MLX library
#2
by l2dy - opened
Unlike https://huggingface.co/unsloth/Qwen3.6-35B-A3B-UD-MLX-3bit, this model (unsloth/gemma-4-26b-a4b-it-UD-MLX-4bit) is not listed on https://huggingface.co/models?library=mlx.
Yeah, their Gemma model cards are all missing a mlx tag and library_name: mlx . I pushed a PR for a separate Gemma version. If you use LM Studio, you can still get and run it, though, by downloading it.
hf download unsloth/gemma-4-26b-a4b-it-UD-MLX-4bit \
--local-dir ~/.lmstudio/models/unsloth/gemma-4-26b-a4b-it-UD-MLX-4bit
For Unsloth team:mlx tag gets it into LM Studio's filterlibrary_name forces HuggingFace to treat it as MLX instead of bare Safetensors.