MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has launched its most advanced artificial intelligence models to date, named Muse Wonder and Muse Wonder Full. These new systems are designed to support coding, long-duration AI agents, and handling complex personal or professional projects. According to Meta, these models can assist users with navigating insurance claims, interpreting leases, analyzing financial data, and developing functional applications. Muse Wonder was made accessible on October 12 via the Meta Model API and Muse Code, while Muse Wonder Full was introduced exclusively for Muse Code subscribers.
This release signifies Meta’s expanded focus on AI systems capable of executing multi-step processes rather than simply responding to isolated prompts. Muse Wonder can plan extended tasks, assign subtasks to sub-agents, utilize software tools, and recover from errors during execution. Meta has also optimized the models for advanced computer use across browsers, desktop apps, and mobile platforms. Both models support a context window of roughly one million tokens, enabling them to process substantial information within a single session.
To facilitate longer and more tailored tasks, Meta introduced two features: Active Memory and Active Focus. Active Memory retains relevant information throughout conversations and retrieves details when they are pertinent to a new task. Active Focus enables the models to maintain progress when facing difficult tasks, test alternative strategies, and recover after errors. Meta emphasized that the models can also seek clarification from users and verify actions before executing significant tasks.
Models Designed for Coding and Persistent AI Agents
Pricing for Muse Wonder is set at $4 per million input tokens and $20 per million output tokens via the Meta Model API. Cached input costs $0.20 per million tokens, as per the company’s developer resources. Muse Wonder Full is only available through paid Muse Code subscriptions, not through separate API pricing. Meta lists Power Usage at $50 a month and Maximum Usage at $200 a month for developers with more intensive coding needs.
Meta reports that Muse Wonder Full outperforms in multiple agentic and coding benchmarks announced alongside the launch. The company highlights tests like DeepSWE and SWE-Marathon, which evaluate software engineering and long-duration coding capabilities. Independent evaluator Artificial Analysis assigned Muse Wonder an Intelligence Index score of 52, and Muse Wonder Full scored 56 at maximum reasoning effort. These results position the models above Meta’s earlier Muse Spark generation on the evaluator’s index.
Enhanced Safety Features for Broader Capabilities
Meta trained the Muse Wonder series to avoid sharing private information without user permission and to decline harmful requests. The models also retain cybersecurity functions that assist developers in identifying vulnerabilities in their code and strengthening defenses. Meta states it assesses its most capable models under the Meta Superintelligence Scaling Framework and plans to publish a comprehensive report on additional safety evaluations following the initial release.
The company began a major overhaul of its AI initiatives approximately 14 months before launching Muse Wonder. Meta indicates that these new models build upon the foundations set by Muse Spark, with training carried out at its Prometheus data center. The plan is to integrate Muse Wonder into more features within its Muse platform, initially targeting subscribers. The October launch provides developers with immediate access to Muse Wonder, while reserving the larger Muse Wonder Full model for Muse Code members.