Meta introduced Muse Wonder and Muse Wonder Full on October 12, describing them as its most capable models to date for personal and professional agentic tasks. The models are designed for coding, computer use and workflows that continue across multiple steps.
The release comes from Meta Superintelligence Labs and builds on the company’s earlier Muse Spark model. Meta says the new systems can handle tasks such as navigating an insurance claim, interpreting a lease, analyzing financial information and building a working application.
Muse Wonder reaches developers through Meta’s API at $4 per million input tokens
Muse Wonder is available through the Meta Model API and Muse Code. Meta lists a context window of 1,048,576 tokens, or roughly 1 million tokens. Standard API pricing is $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens.
Meta’s documentation says the API is compatible with OpenAI SDK-based clients. The documented model identifier is muse-wonder-1.0, and the API base URL is https://api.meta.ai/v1. Meta also lists support for long-running agents, multi-agent workflows, computer use and integrations including OpenRouter.
Muse Wonder Full reserves Meta’s strongest model for Muse Code subscribers
Meta describes Muse Wonder Full as its most advanced model yet. It is available through Muse Code subscriptions rather than the public Model API. Meta’s developer page lists the Power Usage plan at $50 per month and the Maximum Usage plan at $200 per month, with both plans including access to Muse Wonder Full.
The two models include features Meta calls “Active Memory” and “Active Focus.” Active Memory carries relevant information across conversations, while Active Focus is intended to help the model continue working, try different approaches, recover from errors and request clarification when necessary.
Meta’s safety controls target agents that can act outside the chat window
Meta says the Muse Wonder family was trained to avoid sharing private information without permission and to refuse requests that would assist in harmful activity. The company also says the models can help developers find flaws in their own code and strengthen defenses against cyberattacks.
Meta said it trains and evaluates its most capable models under the Meta Superintelligence Scaling Framework. The company also said a fuller report with additional safety evaluations will be available soon.
Try Muse Wonder with a compatible API client
- Create a Meta developer account and obtain a Model API key.
- Configure an OpenAI-compatible client to use
https://api.meta.ai/v1and selectmuse-wonder-1.0. - Begin with a limited coding or research task, then review tool permissions before enabling browser, desktop or other external actions.
Meta’s launch gives developers immediate access to Muse Wonder while keeping Muse Wonder Full inside Muse Code. That split makes the API model available for experimentation and production testing, while the company’s strongest model remains tied to its paid coding plans.
References
Meta. (2026, October 12). Introducing Muse Wonder and Muse Wonder Full: Frontier models for agents. https://about.fb.com/news/2026/10/introducing-muse-wonder-models/
Meta. (2026, October 12). Muse Wonder. https://dev.meta.ai/models/muse-wonder
Meta. (2026, October 12). Get started with Meta Model API and Muse Code. https://dev.meta.ai/docs/overview
Brandom, R. (2026, October 12). Meta releases its first frontier model, Muse Wonder. TechCrunch. https://techcrunch.com/2026/10/12/meta-releases-its-first-frontier-model-muse-wonder/