Muse Wonder is a model for people who need an AI system to carry out extended, multi-step work rather than respond to a single prompt. It is offered through Meta Model API and Muse Code, with capabilities for software engineering, computer-use agents, and interpreting visual material. Its planning, sub-agent delegation, and error recovery are intended to help developers and other users move a task forward across multiple steps, including when the work encounters a blocker.
What can Muse Wonder do?
Muse Wonder can plan long tasks, delegate work to sub-agents, make sequential tool calls, and recover from errors. For software engineering, it can work across large repositories, make multi-file changes and refactors, debug code, and verify work. Computer-use agents can interact with browsers, desktop applications, and mobile apps.
- Interpret screenshots, charts, documents, images, and video alongside conversation.
- Use Active Memory to carry relevant context across conversations.
- Use Active Focus to continue work, try alternatives, and ask for clarification when needed.
- Handle extended multi-step agent tasks and long-horizon execution.
The model has a one-million-token context window. Audio-dependent use cases require confirmation that the particular release surface supports them; support should not be assumed across every way of accessing the model.
Who is Muse Wonder best for?
Muse Wonder is suited to developers handling code across large repositories, multi-file changes, debugging, or refactoring, and to teams building agents that need tool calls, computer interaction, and recovery from errors. It may also suit people whose work depends on interpreting documents, charts, images, or video, or continuing a task with context from earlier conversations. Choose an access route based on whether you need the Model API or Muse Code; Muse Wonder Full is available only through eligible Muse Code subscriptions.
Muse Wonder pricing in brief
Access is available through both subscription and usage-based API routes. Muse Code subscriptions include several usage levels, with higher levels offering larger prompt allowances and, on eligible levels, access to Muse Wonder Full and additional benefits. API use is billed by token, and additional API keys used with subscriptions are also billed pay-as-you-go. These are distinct ways to access the model, so compare the subscription benefits with the API billing route against your expected use before choosing. Subscription benefits can vary by region, and the available information does not establish that every benefit is offered in every location.
Limitations to know about Muse Wonder
Muse Wonder Full is not listed as available through the Model API; it is subscription-only through Muse Code on eligible levels. Muse Wonder and Muse Wonder Full are excluded from Muse Code subscriptions in South Korea, and regional benefits may vary. Subscriptions default to Contributor mode, which permits Meta to use submitted content and responses to train future models; users can turn that mode off, after which future content is excluded from training. Confirm audio support on the specific release surface before relying on audio-dependent workflows.
Alternatives to Muse Wonder
For related AI assistants, ChatGPT is a general-purpose conversational option, while Claude is another assistant to compare for text-based tasks. For workflows organized around multiple collaborating agents, Crew AI is a relevant alternative. Compare the actual tools and access options for your intended work; these alternatives are not presented as having the same features or limits as Muse Wonder.
