Stability AI Unveils 2.7B-Parameter Audio 3.0, Generating Music Beyond 6 Minutes
Updated
Updated · TechCrunch · May 20
Stability AI Unveils 2.7B-Parameter Audio 3.0, Generating Music Beyond 6 Minutes
5 articles · Updated · TechCrunch · May 20
Stability AI launched four Stability Audio 3.0 models, with its top large version generating full music tracks longer than six minutes and the medium and large models reaching 6 minutes 20 seconds.
The release marks a sharp jump from earlier versions: Stable Audio 2.0 more than doubled in length, while 2024’s Stable Audio Open generated up to 47 seconds.
Three models — small SFX, small and medium — are being released with open weights, while the 2.7B-parameter large model is limited to API and paid self-hosting; companies above $1 million in revenue need an enterprise license.
Stability said the new models were trained on fully licensed data, a notable point as AI music rivals face legal scrutiny over training rights and label relationships.
That licensing push builds on Stability’s 2025 deals with Warner Music Group and Universal Music Group and comes as it hires former Universal Audio and Fender executive Ethan Kaplan to lead products for professional musicians.
Can Stability AI's licensed-data model truly outcompete rivals that are using vast, unlicensed datasets?
As AI generates professional music, what becomes the new economic and cultural value of human artists?
Stable Audio 3.0 Sets New Standard for AI-Generated Music: Enterprise Tools, Open-Source Plans, and Industry Impact
Overview
Stability AI officially launched Stable Audio 3.0 on May 20, 2026, marking a major step forward in AI music generation. This release builds on the company’s earlier commitments to creative AI and follows strategic partnerships with Universal Music Group and Warner Music Group, which aimed to develop next-generation professional music creation tools. These partnerships set high expectations for Stable Audio 3.0’s capabilities. The new model family includes a powerful large model, available only through Stability AI’s API and paid self-hosting, signaling a focus on professional and enterprise use. This launch is set to usher in a new era for generative audio.