ModelsNews
Meta ships specialized models to Muse and Meta AI
Meta is rolling specialized versions of its models, trained with Muse Wonder, into Muse and Meta AI. The Personal Superintelligence Index rises from 38% to 47%, with Active Memory and Active Focus for fewer forgotten goals and stronger recovery.
· 2 min read
What shipped?
Meta AI Research said it is bringing specialized versions of its existing models, trained with the Muse Wonder series, to Muse and Meta AI. The goal is wider access to Wonder-class capabilities, not a new open-weight release.
On the Personal Superintelligence Index, the current Muse model scores 47%, up from 38%. That is Meta's own frame for personal and professional agentic tasks.
Active Memory and Active Focus
Active Memory trains models to tell temporary from lasting information and to recall earlier details when they matter again. Meta reports a 2.4× reduction in forgetting critical goal context in Muse, with the Active Memory eval rising from 76% to 90% correct use of critical context.
Active Focus keeps working toward a goal when unexpected issues appear. Meta cites a 67% improvement on everyday tasks with disruptions. The specialized models also multitask better and adapt to mid-run steers without losing the thread.
In Meta AI the same upgrades mean clearer step breakdowns, sharper follow-ups and steadier direction changes mid-chat. For earlier Muse context, see our Muse Spark Open piece.
Key takeaways
- Specialized models roll into Muse and Meta AI, trained with Muse Wonder.
- PSI moves from 38% to 47% on Meta's own index.
- Active Memory: 2.4× less critical forgetting; Active Focus: +67% on disrupted tasks.
Produced with the help of AI, reviewed by an editor. How we make our articles
Reviewed and edited by the Neural Daily desk. About
