Stability AI releases a new audio model that can create six-minute songs

Stability AI released Audio 3.0, their latest music generation model that can create tracks up to six minutes long. That's a meaningful jump from earlier models that maxed out around 30 seconds to a minute.

The interesting part is the small model variant. It runs on-device and generates two-minute tracks, which means you can use it without sending everything to the cloud. For anyone building audio tools or just experimenting with AI music, that local processing matters for speed and privacy.

This fits into the broader pattern we're seeing where AI models are getting both more capable and more accessible. The big versions do impressive things, but the small versions that run locally are often more useful for real work.

Music generation has been heating up lately with tools like Suno and Udio getting traction. Stability's move here is about staying competitive in a space that's quickly becoming crowded. Six-minute generations put them in range of what people actually need for background music, demos, or content creation.

The on-device capability also opens up use cases where cloud-based generation doesn't work well. Think live performances, offline editing workflows, or situations where you need instant results without API latency.

For anyone using AI in creative work, this is another tool in the kit. The quality and licensing terms will determine whether it's actually useful, but the technical specs suggest Stability is taking local deployment seriously.