Quick Answer
Meta Muse Wonder is a new AI model family that Meta says is designed for personal and professional tasks, with a consumer Muse rollout planned for subscribers. Muse Wonder is available now through the Meta Model API and Muse Code, while Wonder Full is initially limited to Muse Code subscribers. Users should treat the consumer integration as planned, not broadly available today.
Key Takeaways
- Meta announced Muse Wonder and Muse Wonder Full on October 12, 2026.
- Muse Wonder has a 1 million-token context window, according to Meta.
- Muse Wonder is available through the Meta Model API and Muse Code now.
- Muse Wonder Full is initially available only through paid Muse Code plans.
- Meta plans to begin a consumer Muse rollout for subscribers, but has not provided a public date.
What did Meta announce with Muse Wonder?
Meta announced Muse Wonder and Muse Wonder Full on October 12, 2026, describing the models as its most capable models to date for personal and professional agentic tasks. Meta gave examples that include navigating an insurance claim, understanding a lease, and building an app. Meta’s announcement presents the models as systems intended to complete work across multiple steps rather than only answer a single question.
Muse Wonder is positioned around agentic work, which means the model is intended to carry out a sequence of tasks with context, tools, and checkpoints. That design could make an AI assistant more useful for projects that involve documents, research, browser actions, and follow-up decisions. The limitation is that Meta’s examples describe intended use cases, not a guarantee that every task will be completed correctly without review.
The practical implication is straightforward: consumers should view Muse Wonder as a developing agent platform rather than a replacement for professional judgment. A model can help organize information from a lease or claim, but users should still verify terms, deadlines, amounts, and legal obligations before making a decision.
Where is Meta Muse Wonder available now?
Meta Muse Wonder is available immediately through the Meta Model API and Muse Code, while Muse Wonder Full is initially limited to Muse Code subscribers. Meta also says that Muse Wonder will soon power experiences in the consumer Muse agent, beginning with subscribers. The staged availability means API users and Muse Code users can access the new model family before most consumer Muse users receive it.
Meta has not provided a public date for the broader consumer Muse rollout. That matters because a planned subscriber rollout is not the same as a feature that every Muse user can enable today. Users who see references to Muse Wonder should check whether the feature appears in their specific Muse interface and account tier before relying on it for ongoing work.
Businesses evaluating the model for connected workflows should also distinguish between API access and a consumer product feature. Companies can use the API for their own software integrations, while the consumer Muse rollout concerns Meta’s own agent experience. Small businesses considering agent connections to business tools may also want to understand how Muse connections for accounting and commerce services affect the data they choose to share.
What do Active Memory and Active Focus do?
Meta says Muse Wonder adds Active Memory and Active Focus, two features intended to support longer-running agent tasks. Active Memory retains relevant context across conversations, while Active Focus is intended to keep an agent working through difficult tasks, recover from errors, and check in before important actions. These features address a common problem with AI assistants: a task can lose important context when it spans many prompts or sessions.
Active Memory could make repeated work more convenient because the model can retain information that remains relevant to a later conversation. Active Focus is intended to add persistence when an agent encounters an obstacle, rather than ending the task after a failed attempt. The benefit depends on what information the model retains, how the user can review that information, and whether the agent correctly identifies an action as important enough to require approval.
Users should be cautious about placing sensitive financial, medical, legal, or identity information into an agent’s persistent context. Memory can improve continuity, but it can also increase the amount of sensitive material associated with an account. The safest approach is to provide only the details needed for a specific task and to review available privacy and memory controls before using the model for personal documents.
How much does Meta Muse Wonder cost?
Meta Muse Wonder is listed at $4 per million input tokens, $0.20 per million cached-input tokens, and $20 per million output tokens through the Meta Model API. Meta lists a 1 million-token context window for the model, which describes the amount of text and other prompt context the system can process within a task. Meta’s model specifications also list the available API and Muse Code access options.
Meta’s Power Usage Muse Code plan costs $50 per month, while the Maximum Usage plan costs $200 per month. Both plans include access to Muse Wonder Full. These prices matter most for developers and heavy users because agentic tasks can use substantial input and output processing when they include long documents, repeated tool calls, or extended work sessions.
| Access option | Availability | Listed price | Best fit |
|---|---|---|---|
| Muse Wonder API input | Available now | $4 per million input tokens | Applications sending new prompt material |
| Muse Wonder API cached input | Available now | $0.20 per million cached-input tokens | Applications reusing prior context |
| Muse Wonder API output | Available now | $20 per million output tokens | Applications generating model responses |
| Muse Code Power Usage | Available now | $50 per month | Subscribers who need Wonder Full access |
| Muse Code Maximum Usage | Available now | $200 per month | Subscribers with higher usage needs |
Meta’s listed prices are useful for estimating access costs, but total spending depends on how much context and output an application uses. Developers should measure token use during a limited test before committing to a production workflow, especially when a task includes large files or lengthy agent sessions.
What kinds of tasks can Muse Wonder agents handle?
Meta says Muse Wonder can power agents that operate browsers, desktop apps, and mobile apps, and can hand off multi-step work that runs for hours. The company’s examples include insurance claims, lease review, and app building, which are tasks that may involve reading information, moving between services, and waiting for later steps. The central promise is not only language generation, but coordinated work across software tools.
Long-running agent work can be useful when a task has a clear goal and a user can review key choices. For example, an agent could gather relevant material, organize a draft, and ask for confirmation before an important action. The practical limit is that an AI system can misunderstand an instruction, choose an incomplete source, or apply the wrong context even when it continues working for an extended period.
Users should keep approval requirements in place for purchases, account changes, document submissions, and other consequential actions. The need for review becomes more important when an agent can operate a browser or app on the user’s behalf. Consumers evaluating similar persistent systems can also consider how long-running AI agents change the balance between automation and user oversight.
What are the privacy and safety limits of Muse Wonder?
Meta Muse Wonder should not be used as an unsupervised decision-maker for legal, insurance, financial, medical, or identity-sensitive matters. Meta says Active Focus is designed to check in before important actions, but the company has not yet released its full report with additional safety evaluations for the Wonder models. That gap matters because agent systems can take actions across browsers and apps, where an incorrect step can have consequences beyond a wrong text response.
Users should review every important recommendation, verify facts against original documents, and avoid granting broader account access than a task requires. A model that reads a lease can help identify questions, but it cannot determine a user’s legal rights. A model that helps navigate an insurance claim can organize information, but it cannot guarantee coverage or replace an insurer’s official process.
Privacy risks also increase when an agent has access to messages, documents, business accounts, browser sessions, or saved payment details. The sensible approach is to start with low-risk tasks and remove access that is no longer necessary. Reports involving a Muse-related address incident also show why users should be careful about what personal details an agent can retrieve, infer, or act upon.
Will Meta open-source the Muse Wonder models?
Meta has not said whether it will open-source the Muse Wonder models. The company announced API and Muse Code access, but did not announce a public release of model weights in its launch materials. Axios reported that Meta has not disclosed whether the Wonder models will be open-sourced, while the company is presenting the launch as evidence that it can compete with leading AI labs. Axios’s reporting identifies the unanswered question around the models’ future distribution.
The distinction matters because API access, subscription access, and open-weight availability give developers different levels of control. API access lets developers build on a hosted model, while an open-weight release would allow qualified organizations to run and adapt the model in their own environments. Meta has not committed publicly to that latter option for Muse Wonder.
Developers should make plans based on the access methods Meta has actually announced rather than assuming a future open release. The current practical choice is between using Muse Wonder through Meta’s API or accessing Muse Wonder Full through eligible Muse Code subscriptions.
Who should wait for the consumer Muse Wonder rollout?
Most consumer users should wait for Meta to make Muse Wonder available in the consumer Muse agent before expecting access to its new agent capabilities. Meta says the rollout will begin with subscribers, but the company has not announced a public schedule for broader availability. That limitation means users should not assume that a Muse account already includes Muse Wonder or Muse Wonder Full.
Developers, businesses, and advanced users who need immediate access have clearer options because Muse Wonder is available through the Meta Model API and Muse Code. Those users should still begin with contained tasks that have clear review points, particularly when the model can work across browser, desktop, or mobile environments. A staged test can reveal whether the model’s context handling and cost structure fit a real workflow.
For most people, the practical response is to wait for the subscriber rollout, check the exact account availability, and avoid moving sensitive work into an agent until its controls and behavior are clear. Muse Wonder may make multi-step tasks easier to manage, but users remain responsible for reviewing the information and actions that matter.
FAQ
What is Meta Muse Wonder?
Meta Muse Wonder is a new AI model family that Meta says is built for personal and professional agentic tasks. Meta describes uses that include working through insurance claims, lease information, and app-building tasks.
Is Meta Muse Wonder available now?
Meta Muse Wonder is available now through the Meta Model API and Muse Code. Muse Wonder Full is initially limited to Muse Code subscribers, while consumer Muse availability is planned to begin with subscribers.
How much does Meta Muse Wonder cost?
Meta lists Muse Wonder API pricing at $4 per million input tokens, $0.20 per million cached-input tokens, and $20 per million output tokens. Muse Code plans with Wonder Full access cost $50 per month for Power Usage and $200 per month for Maximum Usage.
Does Muse Wonder remember previous conversations?
Meta says Muse Wonder includes Active Memory, which retains relevant context across conversations. Users should avoid placing unnecessary sensitive information into persistent AI context and should review available privacy settings.
Can Muse Wonder take actions in apps and browsers?
Meta says Muse Wonder can power agents that operate browsers, desktop apps, and mobile apps. Users should keep human approval in place for purchases, account changes, submissions, and other consequential actions.