4 ms·
https://huggingface.co/lmstudio-community/gemma-4-26B-A4B-it-QAT-MLX-4bit https://huggingface.co/lmstudio-community/gemma-4-26B-A4B-it...
by int_19h 5mo ago
https://huggingface.co/lmstudio-community/gemma-4-26B-A4B-it-QAT-MLX-4bit https://huggingface.co/lmstudio-community/gemma-4-26B-A4B-it...
- dofm 5mo agoYeah — that is the base QAT model, and there are safetensors weights for the QAT version of the MTP drafter, but there are no MLX/GGUF versions. I think the answer is a combination of: 1) Gemma 4 MTP is too fresh for off-the-shelf software to use anyway 2) "you can convert them yourself" which is fine, obvs