Stability AI has unveiled Stable Audio 3.0, a new family of audio models that can generate tracks lasting 6 minutes 20 seconds, TechCrunch reported on Tuesday.

The company said Stable Audio 3.0 consists of Small SFX, Small, Medium and Large, and the top-tier model supports professional-grade music creation that runs longer than 6 minutes.

Small SFX and Small each have 459 million parameters and support on-device generation of sounds and music of up to 2 minutes. Medium has 1.4 billion parameters and Large has 2.7 billion.

Both the Medium and Large models can produce complete tracks lasting 6 minutes 20 seconds. They can also maintain musical structure and melodic tone, the company said. That is more than twice as long as Stable Audio 2.0, unveiled in 2024.

Stability AI released the weights for Small SFX, Small and Medium so anyone can use and modify them. Large, by contrast, is available only through an API and a paid self-hosted service. Companies with annual revenue exceeding $1 million must obtain a separate enterprise licence.

Stability AI said this family of audio models was built on data for which it had secured full licences. Stability AI also signed agreements last year with Warner Music Group and Universal Music Group to develop models and music production tools.

Stability AI is also developing a new product lineup for professional musicians but did not disclose specific features. Ethan Kaplan, a former chief digital officer at Universal Audio and Fender, has joined Stability AI and is leading development of professional music products, TechCrunch reported.