Stability AI launches Stable Audio 3.0 open-weight music model family

Details

• Stability AI announced Stable Audio 3.0, a new open-weight model family for generative audio aimed at artistic experimentation.
• The company says users own the outputs and can distribute the generated audio.
• The release includes four models tuned for different use cases and deployment options.
• Stable Audio 3.0 uses a new semantic-acoustic autoencoder designed to produce longer and more flexible audio generations.
• The models are available through Hugging Face for Small and Medium, while Large can be used via Stability AI’s API or self-hosted with an enterprise license.
• Stability AI also opened a waitlist for an upcoming product suite for musicians, signaling a broader commercial push around the platform.

Impact

Open-weight audio models can lower adoption barriers for creators, startups, and enterprise teams that want more control over cost, deployment, and rights management. By pairing downloadable smaller models with an API and self-hosting path for the largest version, Stability AI is pushing a more flexible distribution strategy than many closed music-generation tools. The move may pressure rivals in AI audio to explain their licensing terms, output ownership, and deployment options more clearly.