MiniMax Releases Open-Weights Music Model Music 3, Gets Day 0 Support from SGLang

ying11231 · x · 2026-08-14

MiniMax has released MiniMax-Music 3, a next-generation, production-ready, and versatile open-weights music generation model.

The model is powered by a Qwen3 + RVQ autoregressive backbone and a flow-matching DIT/DAV decoder. SGLang has announced Day 0 support for the model, enabling developers to deploy and test it quickly. MiniMax also released a demo video featuring parkour in old Shanghai alleys, combining visuals from Hailuo AI with music generated by Music 3.

Related event: MiniMax Open-Sources Music 3 Text-to-Music Model(14 posts)→

Original post →

More from Multimodal

Multimodal channel →