Meta is rolling out specialized versions of the models behind Muse and Meta AI, trained with Muse Wonder to bring some of the flagship series’ capabilities to a wider group of users. The move follows the introduction of Muse Wonder a few weeks ago, which Meta described as its most capable model series yet, with state-of-the-art results on personal and professional agentic tasks. The new models are now being introduced for people who use Muse and Meta AI.
Upgraded the models for Muse and Meta Al, bringing Active Memory and Active Focus from Muse Wonder to everyone. In Muse, these new specialized models are 2.4x less likely to forget critical info and are 67% better at self correcting and completing everyday tasks which means less…
— Mark Zuckerberg (@finkd) November 6, 2026
The rollout brings Active Memory and Active Focus, two systems first introduced with Muse Wonder, to the specialized models. Active Memory allows Muse and Meta AI to retain what matters and recognize when details from earlier conversations become relevant again. The models were trained to separate temporary information from permanent information, reason across time, and identify what remains useful or has become stale. Meta says its current Muse models show a 2.4x reduction in forgetting critical information needed to help users reach their goals.
1/ starting today, muse and meta ai will run on new specialized models trained using muse wonder.
— Alexandr Wang (@alexandr_wang) November 6, 2026
ACTIVE MEMORY: models can reason across time and better remember what matters
ACTIVE FOCUS: enabling muse to continue working towards a goal even when met with roadblocks pic.twitter.com/7gt9vbGEvb
On Meta’s Active Memory evaluation, current Muse models correctly use critical context in 90% of tasks, compared with 76% for previous models. They also score 47% on the Personal Superintelligence Index, up from 38% for the earlier generation.
Active Focus is designed to keep Muse working toward a goal when unexpected problems appear. Training covered bottlenecks modeled on real-world situations, teaching the models to persist, recover, or find another route on a user’s behalf. Meta reports a 67% gain in completing everyday tasks where unexpected issues arise.
The specialized Muse models also draw on Muse Wonder’s performance in behaviors that matter for agentic work. They can handle multiple tasks, decide when to pursue goals independently or collaborate live with a user, and adapt to steers or clarifications while a task is running without losing track of the objective.
In Meta AI, the same work is intended to create a stronger thinking partner and collaborator. It can divide complex problems into clear steps, ask sharper follow-up questions, and stay with the conversation when a user changes direction. Meta says the broader goal is to give more people access to more capable superintelligence by transferring key Muse Wonder capabilities into the services they already use.
Sources and related context
- Muse Wonder introduces Active Memory and Active Focus: Related context: Meta’s earlier Muse Wonder announcement explains Active Memory and Active Focus and outlines plans to use Wonder to develop specialized models for Muse.
