Music generation AI “Stable Audio 3.0” debuts, with a large model that can generate over 6 minutes of music and a small model that runs on smartphones

Stability AI released Stable Audio 3.0, a four-variant music generation family spanning open models and a high-end large model.

The small models can run on smartphones and laptops, while the large model can generate or extend music up to 6 minutes 20 seconds.

The system also supports editing existing audio, continuing tracks, and LoRA fine-tuning.

Small and medium models are available on Hugging Face, while large-model access is offered via API and enterprise channels.