MiniMax Music 3 Open-Weights: Prompt-to-Song Generation on Consumer GPUs

multimodalart · x · 2026-08-14

MiniMax has released Music 3 with open weights on Hugging Face. The model utilizes an 8B LLM combined with a 2.7B DiT architecture to generate full songs from text prompts and lyrics.

It is optimized to run on consumer-grade GPUs via 🧨 Diffusers and ComfyUI. An interactive demo is also available on Hugging Face Spaces.

Related event: MiniMax Releases and Open-Sources Music 3 Generation Model(10 posts)→

Original post →

More from Models

Models channel →