Stability AI is releasing a new family of audio models called Stable Audio 3.0, with a top-tier 2.7 billion-parameter version capable of generating professional-grade music lasting more than six minutes. The release includes four models of varying sizes, three of which will be available with open weights for users to download, modify and build upon.

The lineup consists of a small SFX model and a small model, each with 459 million parameters, a medium model with 1.4 billion parameters, and the large model with 2.7 billion parameters. The small SFX and small models support on-device sound and music generation of up to two minutes, while the medium and large models can create full compositions of up to six minutes and twenty seconds while maintaining musical structure and melodic tone.

The maximum composition length is more than double what Stable Audio 2.0, released in 2024, was capable of producing. Also in 2024, the company released Stable Audio Open, which allowed for music generation of up to 47 seconds.

The small SFX, small and medium models are being made available with open weights for anyone to use and modify. The large model is available only through the API and self-hosting paid services. Companies with annual revenue over $1 million must obtain an enterprise license to use the large model.

A spokesperson said the latest set of audio models is built on fully licensed data. Last year, Stability AI signed deals with Warner Music Group and Universal Music Group to develop audio models and music-creation tools.

Stability AI is also developing a new suite of products aimed at professional musicians. Ethan Kaplan, former chief digital officer at Universal Audio and Fender, joined the company to lead its professional music offering. Most of the new Stable Audio 3.0 offerings are free to download and build upon, allowing users to customize their own models.