The current Muse scores 47% on Meta – Unite.AI’s personal superintelligence index

Meta AI Research announced on November 6, 2026 that it is launching specialized versions of the models behind Muse and Meta AI, trained using Muse Wonder, with the current Muse model scoring 47% on its Personal Superintelligence Index compared to 38% for the previous model.

The update brings Active Memory and Active Focus to both products. The models implemented are specialized versions of Meta’s existing systems, improved through Muse Wonder, rather than a launch of the Wonder series itself. This approach extends the features introduced with Wonder to more people through existing assistant models.

In its October 12, 2026 announcement, Meta described Muse Wonder and Muse Wonder Full as its most capable models to date, and said Wonder will power Muse experiences for subscribers before being used to develop specialized models for everyone using Muse. The new implementation brings that specialized model plane to both Muse and Meta AI.

Active memory selects the relevant context

Active memory allows assistants to retain important information and retrieve it when it becomes relevant for a later conversation. The training covers the difference between temporary and permanent information, reasoning over time and recognizing when details remain relevant or have become obsolete. This makes memory a selection and timing problem: the model must identify which previous context applies now and which has become obsolete.

For the improved Muse models, Meta reports a 2.4x reduction in forgetting information critical to users’ goals. Its active memory assessment shows the current model correctly using the critical context in 90% of tasks, compared to 76% in the previous model. These figures describe separate parts of the problem: retaining the information needed for a goal and correctly applying that context during a task.

Active focus keeps things moving

Active Focus is intended to keep Muse moving toward a goal when an unexpected challenge interrupts work. Meta trained models on various bottlenecks designed to reflect real-world conditions so they can persist, recover, or pursue alternatives for the user. In Muse, Meta reports a 67% improvement in completing daily tasks where unexpected problems occur.

Specialized patterns also address how an agent collaborates with someone while a task is in progress. They are designed for multitasking, choose between independent progress and live collaboration, and incorporate guidance or clarification without getting confused or missing the task. This links persistence to responsiveness: an agent can continue to pursue the goal while still changing course when the user intervenes.

In Meta AI, the same improvements are intended to make the assistant a smarter partner and collaborator. It can break down complex problems into clear steps, ask more precise follow-up questions, and keep up when a user changes direction during a conversation.

For users, the practical test of launching is continuity: if the useful context survives changing circumstances, the work recovers from obstacles and fixes can be incorporated without losing the goal. These behaviors connect the results reported by Meta to the daily collaboration it intends to support in Muse and Meta AI.