Meta releases Muse Wonder and Muse Wonder Full, its first frontier AI models

Meta said Muse Wonder is available through the Meta Model API and in Muse Code, priced at $4 per million input tokens and $20 per million output tokens. Reports differed on subscription access. Engadget reported that Muse Wonder Full is limited to Muse Code subscribers on $50-a-month “power usage” or $200-a-month “maximum usage” tiers. CNBC reported that users of the Muse personal agent app need a $20 or $100 monthly subscription for Muse Wonder, with Muse Wonder Full restricted to Muse Code subscribers on $50 or $200 plans. TechCrunch reported that Wonder Full is not served through the API. The Verge said both models are available for Muse Code subscribers starting today and that Wonder can also be accessed through the Meta Model API.

Meta says internal testing on several industry benchmarks shows the models achieve comparable scores to OpenAI's GPT-6 Astra and Anthropic's Claude Fable 5.1 and Claude Opus 5.5. The Verge reported it is unclear how they compare with Google's Gemini 4 Argon. Meta says the models are particularly adept at legal, financial and health-related tasks and deliver “frontier coding performance.” They add “active memory” and “active focus,” which Meta describes as carrying what matters to a user across conversations and keeping an agent working toward a goal when a task gets hard.

According to Meta's testing, Muse Wonder scores higher on the MemoryBench benchmark than Fable, Opus or Astra. “We put a tremendous amount of effort into making sure your model can remember what’s important to you over days, weeks, months and beyond,” research scientist Kavya Kopparapu said in a video accompanying the launch.

Meta plans to eventually power its Muse AI agent with Wonder, starting with subscribers, and to develop specialized models it “can bring to everyone using Muse,” according to The Verge. It has not shared specifics or a timeline. Meta AI chief Alexandr Wang said the model, codenamed Watermelon, was developed to improve “personal and professional agents” so they do not stall on real-world tasks such as booking dinner reservations and checking for availabilities. “When you do real-world tasks, you’re going to get roadblocks and you’re going to get things that come up or various setbacks,” Wang said. “We specifically tried to train the model to be able to try other approaches instead of stalling and then checking back in with the user at any time.” Meta said the model offers its strongest safeguards to date against “cyber misuse,” and those protections come only with the paid model for now.

Muse shot to the top of Apple's App Store after Meta introduced the AI agent app last month. Some users have complained on social media that the agent accessed personal messages without authorization, which the company has disputed. Wang said Meta's AI team has been monitoring for agent-related “incidents or issues” with Muse. “I think there are certainly instances users have authorized various communications or whatnot earlier in a conversation, and then may have forgotten about that,” he said, adding that Meta is educating users “to ensure that consumers can trust the product going forward.”

The release is a milestone in Meta's aggressive push to reestablish itself in AI. The company has poured billions into talent and infrastructure and expects to spend as much as $145 billion this year. CEO Mark Zuckerberg told the Financial Times that “getting to this point both required a large amount of compute and investment, and a lot of very rigorous scientific work.” The models were trained at Meta's Prometheus data center in central Ohio, which came online earlier this year. CNBC reported that Meta's $14.3 billion investment in data-labeling startup Scale AI is starting to pay off through new AI products. TechCrunch noted the release is the first frontier model since Scale AI founder Alexandr Wang restructured Meta's AI efforts in August last year.

The launch came as the AI-driven bull market marked its fourth anniversary on Monday. The S&P 500 has climbed nearly 120% since Oct. 12, 2022, and topped 7,800 for the first time last week, though investors are weighing inflation, bond yields and a more hawkish Federal Reserve, CNBC reported. Options volume in Snap rose to more than four times its month-long average on Monday as excitement swirled around Meta's AI agent, according to CNBC; Snap shares are up almost 20% this month.