Meta is introducing the Muse Wonder series with Muse Wonder and Muse Wonder Full on October 12, 2026. Muse Wonder is available today through the Meta Model API and Muse Code, according to Meta.
The Essentials
- Muse Wonder will soon power more of Muse, rolling out first to subscribers, per Meta.
- Muse app access requires a monthly subscription of $20 or $100.
- Meta says it trained the Muse Wonder family to avoid sharing private information without permission.
- Some users say the agent read their personal messages without authorization. Meta has disputed that.
Meta calls the pair our most capable models to date. It also says they deliver state-of-the-art performance on personal and professional agentic tasks.
Those are Meta’s own descriptions. The Meta newsroom post cites no benchmark figures.
Alexandr Wang, Meta’s AI chief, said in an interview on Monday that Meta built Muse Wonder, codenamed Watermelon. He said the goal was to improve “personal and professional agents.”
The post adds that the models have Active Focus, the ability to keep working toward your goal when a task gets hard.
Introducing Muse Wonder, our new family of frontier models with state of the art performance for personal and professional agentic work. https://t.co/NKHrybrGKQ
— Mark Zuckerberg (@finkd) October 12, 2026
Prices split Muse Wonder across app, API and Muse Code
Access to the models splits across the Muse app, the API and Muse Code. Meta’s own post lists no prices, so every figure here rests on a single news report. API access, Wang said, will cost more than the standard subscriptions, yet stays “meaningfully below” leading models from OpenAI and Anthropic.
Muse Wonder Full will only be available to developers who sign up to the company’s software engineering focused Muse Code AI agent. Access comes under a $50 or $200 a month subscription plan. That keeps the even more powerful variant inside Meta’s coding agent rather than its consumer app.
Permission complaints test the privacy pitch
Meta’s privacy case rests on training and a check-in habit. The post says a personal agent often means working across your messages, calendars, and documents. It adds that Active Focus involves checking in before following through on anything important.
Some users have complained on social media that the agent accessed personal messages without authorization. Wang’s explanation was narrow. I think there are certainly instances users have authorized various communications or whatnot earlier in a conversation, Wang said. Those users, he added, may have forgotten about that.
Cyber safeguards are tiered by price. Meta said in a blog post that the model “offers our strongest safeguards to date” against “cyber misuse”. Those protections only come with the paid model for now.
The post also says developers and security teams can use models like Muse Wonder to find flaws in their own code.
The dispute turns on how long an approval lasts. Safeguards of this kind help reduce risk, but they do not settle what a user approved earlier.
Meta says a full report with additional safety evaluations will be available soon. It gives no date, so for now the privacy and cyber claims rest on Meta’s own account.
