Meta Brings Specialized Muse Wonder Models to Muse and Meta AI
Today we're bringing specialized models trained using Muse Wonder to @Muse and Meta AI, giving users access to some of Muse Wonder's frontier capabilities for complex tasks at work and in everyday life. These specialized models include Muse Wonder's Active Memory and Active Focus, helping them stay aligned with your goals as tasks unfold: → Active Memory helps recognize when earlier information matters and distinguish what's still relevant from what's become stale → Active Focus helps recover from setbacks or find alternatives when constraints change In Muse, the improved models show a 2.4x reduction in forgetting information critical to your goals and a 67% improvement in getting unstuck as constraints evolve. That means less repeating yourself and fewer dead ends when plans change. In Meta AI, these improvements help it break complex problems into steps, ask sharper follow-up questions and adapt when you change direction. Read more on our research blog:
Context
Meta AI and Muse users get models that Meta says forget 2.4x less and get unstuck 67% more often.
Muse Wonder is Meta Superintelligence Labs' frontier model family. Meta says the new specialized models carry two of its capabilities, Active Memory and Active Focus, into the Muse app and Meta AI. The same day Alexandr Wang announced Spaces in Muse, pages Muse builds and keeps up to date from a chat request.
Deep dive
AI at Meta said on Friday that Muse and Meta AI now run on specialized models trained using Muse Wonder, its frontier model family. The goal, according to Meta, is to bring some of Muse Wonder's capabilities for long, complex tasks to everyday users rather than keeping them in the flagship model.
Meta highlighted two capabilities. Active Memory is meant to recognize when earlier information in a conversation still matters and to tell relevant details apart from stale ones. Active Focus is meant to help the model recover from setbacks or find alternatives when constraints change during a task.
Meta gave two numbers for Muse: a 2.4x reduction in forgetting information critical to the user's goals, and a 67% improvement in getting unstuck as constraints evolve. In Meta AI, the company says the update helps the assistant break complex problems into steps, ask sharper follow up questions and adapt when the user changes direction. These figures come from Meta's own testing and have not been independently measured.
Alongside the model update, Alexandr Wang announced Spaces in Muse. A user describes what they want in chat, Muse builds a page on that topic and keeps it current, and the page can be pinned to the navigation bar.
What to expect next
Watch for independent tests of the memory and focus claims, and for whether Meta publishes benchmark results for the specialized models beyond its own blog figures.
Open questions include which Muse Wonder capabilities stay exclusive to the flagship model, how fast the update reaches every Meta AI surface, and how Spaces compares with similar live pages from OpenAI and Google. Usage numbers for Muse and Meta AI after the update would show whether the changes bring people back.