Voxtral Mini 4B Realtime 2602 is published by Mistral AI under Apache-2.0.
https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602

mlx-community converted and quantized the model to 6-bit MLX using voxmlx:
https://huggingface.co/mlx-community/Voxtral-Mini-4B-Realtime-6bit

Noted's Voxtral encoder, decoder and audio adapter are Swift ports of voxmlx:
https://github.com/awni/voxmlx
Copyright (c) 2026 Awni Hannun. MIT License (included as voxmlx-MIT.txt).
The Python model card's awnihannun/voxmlx link is stale; awni/voxmlx is the
verified public source. Noted does not modify the downloaded model weights.

The full Apache-2.0 license and upstream model cards are included.