Has anyone tried this for Nemotron 3 Nano (MXFP4)?

[TomLucidor]

I am kinda shopping around for the best way to run MLX with 4-bit

• https://huggingface.co/noctrex/Nemotron-3-Nano-30B-A3B-MXFP4_MOE-GGUF/tree/main
• https://huggingface.co/ExaltedSlayer/nvidia-nemotron-3-nano-30b-a3b-mlx-mxfp4/tree/main
• https://huggingface.co/mlx-community/NVIDIA-Nemotron-3-Nano-30B-A3B-MLX-MXFP4/tree/main
• Maybe? https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3-Nano-30B-A3B-MLX-4bit

This seems to be one OpenAI-compatible servers out of many, and some of them are already outdated. Since Nemotron relies on Mamba2, would be good to see if MLX-OpenAI-Server to be as good as MLX-Omni-Server (they didn't update their version for 2 months). musistudio/claude-code-router#512 (comment)