Meta brings Muse Wonder–trained specialized models to Muse and Meta AI
Meta introduced specialized versions of its existing models trained with Muse Wonder and is rolling them out in Muse and Meta AI, including Active Memory and Active Focus. Meta says Muse shows a 2.4× reduction in forgetting critical goal information and a 67% improvement on everyday tasks when unexpected issues arise.
Meta announced on November 6, 2026 (Singapore time) on its research blog that it is introducing specialized versions of its existing models trained using Muse Wonder, and rolling them out to people who use Muse and Meta AI.
A few weeks earlier Meta introduced the Muse Wonder series. This update brings some of Wonder’s frontier capabilities—including Active Memory and Active Focus—to the models behind Muse and Meta AI for a wider set of users.
According to Meta, Active Memory helps models recognize what information is temporary versus lasting, and distinguish what is still relevant from what has become stale. Meta says the improved Muse models show a 2.4× reduction in forgetting information critical to a user’s goals; on an Active Memory eval, the current Muse model correctly used critical context on 90% of tasks versus 76% previously. Meta also reports a Personal Superintelligence Index score of 47% for the current Muse model versus 38% previously.
Active Focus helps models keep working toward a goal when unexpected challenges appear—trying harder, recovering, or finding alternatives. Meta says Muse models show a 67% improvement completing everyday tasks where unexpected issues arise. The specialized models are also described as stronger multi-taskers that know when to push goals independently or collaborate live, and that adapt better to mid-task steers. In Meta AI, Meta says the same improvements help break complex problems into steps, ask sharper follow-ups, and keep up when users change direction.
Official post: Bringing Specialized Models to Muse and Meta AI
Official account post: AIatMeta