Meta announced Muse Wonder and Muse Wonder Full on October 12, 2026. Meta’s documentation lists Wonder for Model API in Standard mode and Muse Code, while Wonder Full is not offered through the API and is available in two Muse Code plans. Meta’s announcement and developer documentation set out the distinction.
What Active Memory and Active Focus are designed to do
Meta describes Active Memory as carrying useful context between conversations and retrieving it for the task at hand. Active Focus is the model’s ability to keep working toward a difficult goal, try alternatives, adapt to new information, ask for clarification when needed and check in before important actions. What a model can actually do also depends on the tools and controls connected to it. Meta’s announcement says Muse Wonder is intended for personal and professional agent tasks and that it would soon power experiences in the Muse agent, initially for subscribers.
Meta’s developer documentation lists text, image, video, audio and PDF inputs for Muse Wonder, with text output and a context window of 1,048,576 tokens. Its reasoning-effort setting can range from minimal to max; choosing max does not select Muse Wonder Full. Meta’s model documentation identifies Full as a separate model variant.
The model variant determines the access route
Meta’s model documentation and Muse Code subscription page list these routes:
| Model | Meta Model API | Muse Code |
| Muse Wonder | Standard mode (muse-wonder-1.0) | Pay-as-you-go and all subscription plans, in Standard or Contributor mode |
| Muse Wonder Full | Not offered | Power Usage and Maximum Usage plans, in Standard or Contributor mode |
That split matters if you are choosing a development path: the API route is for Muse Wonder, while access to Wonder Full requires one of the two named Muse Code plans. The max reasoning setting on Wonder does not change which model you are using.
API rates and training settings depend on the mode
In Standard mode, Meta lists rates of $4 per 1 million input tokens, $0.20 per 1 million cached input tokens and $20 per 1 million output tokens. Web-search grounding adds $2.50 per 1,000 queries. These are Meta Model API rates. Meta’s pricing documentation gives the rates and the separate search charge.
Meta says Model API Standard mode does not use submitted prompts, code, files, context or model responses to train its models. Muse Code subscriptions default to Contributor mode, which allows submitted content to be used to train future Meta models; subscribers can turn that setting off for future content. Meta’s pricing and Muse Code subscription documentation describe the two modes.
Artificial Analysis shows a dated leaderboard snapshot
On October 13, Artificial Analysis listed Muse Wonder Full (Max) at 56 points, fourth among 689 models in its comparison class. Artificial Analysis’s model page displays that score and rank. The day before, Trending Topics reported an earlier Artificial Analysis Intelligence Index snapshot that placed Wonder Full at 55.8 points and second. Trending Topics’ October 12 report described that earlier ranking; the two positions come from separate dated snapshots.
