Stability AI Releases Stable Audio 3.0 — Open-Weight Models Generate 6-Minute Songs On-Device

TechCrunch·medium signal

Stability AI released four Stable Audio 3.0 models (459M to 2.7B parameters) capable of generating professional-grade music up to 6 minutes 20 seconds — more than double the length of Stable Audio 2.0. The small and medium models ship with open weights for anyone to modify; the 2.7B large model is API/enterprise only. On-device capable small models generate 2-minute tracks, while medium and large maintain musical structure and melodic tone over full compositions. Companies above $1M revenue need an enterprise license.