Meta Says Muse Is About to Get a Lot Smarter and Safer With New AI Model

Meta Muse Wonder AI model

Meta just announced two new AI models, Muse Wonder and Muse Wonder Full, built to handle more complicated tasks and remember what you’ve told it across conversations.

Meta unveiled both on Monday and calls them its most capable models yet. They’re aimed at AI agents for personal and work use, which means getting things done for you instead of just answering questions.

Muse Wonder is available today to developers through the Meta Model API and Muse Code, and Muse Wonder Full is available to Muse Code subscribers. Meta says Muse Wonder will power its Muse app down the road, starting with paying subscribers. Meta didn’t say when that will happen or whether Canada will get it at the same time.

The headline feature is Active Memory, which lets Muse remember important details from earlier chats and bring them up when they matter. Another one, Active Focus, keeps the model working on longer tasks and gets it to try a different approach when something goes wrong.

Meta says coding is better too, including bigger software projects that take several rounds of testing and fixes. In one demo, Muse Wonder worked across apps and a web browser to build a personalized running plan.

Meta also shared benchmark results, including scores from the Artificial Analysis Intelligence Index, which rolls a bunch of AI tests into one overall number. Muse Wonder Full scored 56, putting it second behind Claude Opus 5.5 at 58 and ahead of Claude Fable 5.1, GPT-6 Astra and Gemini 4 Argon, which all landed at 53.

Artificial Analysis Intelligence Index chart showing Muse Wonder Full at 56 and Muse Wonder at 52.

The regular Muse Wonder model scored 52, tying GPT-6.1 Sol. That’s a big jump from the original Muse Spark at 31, and Meta’s chart shows each Muse Spark update climbing from 34 to 40 to 48 before Muse Wonder. Benchmarks don’t always match how an AI feels to use day to day, so we’ll see how it holds up once people get their hands on it.

Both models were trained at Meta’s Prometheus data centre. Meta says it has added stronger safeguards against misuse and the mishandling of private information, with a more detailed safety report coming later.

Meta keeps pushing Muse beyond a regular chatbot, toward an assistant that can actually finish things for you. Would you trust it to remember your stuff?

Want to see more of our stories on Google?

Add iPhone in Canada as a Preferred Source on Google

P.S. Want to keep this site truly independent? Support us by buying us a beer, treating us to a coffee, or shopping through Amazon here. Links in this post are affiliate links, so we earn a tiny commission at no charge to you. Thanks for supporting independent Canadian media!

Subscribe
Notify of
guest
0 Comments
Oldest
Newest Most Voted
0
Would love your thoughts, please comment.x
()
x