MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has launched its most advanced artificial intelligence models to date, named Muse Wonder and Muse Wonder Full. These systems were specifically engineered for tasks involving coding, extended AI agents, and complex personal or professional activities. According to Meta, the models are capable of assisting users in managing insurance claims, interpreting leases, analyzing financial data, and developing functional applications. Muse Wonder became accessible on October 12 via the Meta Model API and Muse Code, with Muse Wonder Full made available to Muse Code subscribers.
This launch marks an expansion of Meta’s focus on AI systems capable of executing multi-step tasks, rather than solely responding to individual prompts. Muse Wonder can plan lengthy projects, assign subtasks to sub-agents, utilize software tools, and recover from errors during processing. Additionally, Meta designed these models for advanced computing tasks across browsers, desktop applications, and mobile platforms. Both models feature a context window of roughly one million tokens, enabling them to process extensive data within a single session.
To support more extended and personalized interactions, Meta integrated two features called Active Memory and Active Focus. Active Memory maintains relevant information across conversations and retrieves details when they are pertinent to a new task. Active Focus helps the models persist when a task becomes challenging, test alternative approaches, and recover from errors. Meta also states that the models can seek clarification from users and verify actions before executing critical tasks.
Models focus on coding and long-duration AI agents
Pricing for Muse Wonder is set at $4 per million input tokens and $20 per million output tokens through the Meta Model API. Cached input costs are $0.20 per million tokens, according to the company’s developer resources. Muse Wonder Full is only available through paid Muse Code subscriptions, not separate API options. Developers with more intensive coding needs can opt for Power Usage at $50 per month or Maximum Usage at $200 per month.
Meta reports that Muse Wonder Full outperforms many benchmarks related to agentic and coding tasks. The company highlights tests like DeepSWE and SWE-Marathon, used to evaluate software engineering and long-running coding performance. Independent evaluator Artificial Analysis assigned Muse Wonder an Intelligence Index score of 52 and Muse Wonder Full a score of 56 at maximum reasoning effort. These scores indicate the models surpass Meta’s earlier Muse Spark generation on the evaluator’s index.
Enhanced safety measures accompany expanded model capabilities
Meta has trained the Muse Wonder series to prevent sharing private information without consent and to decline harmful requests. The models also include cybersecurity functions that assist developers in identifying vulnerabilities in their code and strengthening security defenses. Meta states that its most capable models are assessed under the Meta Superintelligence Scaling Framework, and plans to publish a comprehensive report on additional safety evaluations following the initial release.
The development of these models followed a broad overhaul of Meta’s AI efforts approximately 14 months prior to the Muse Wonder launch. The models build upon the foundation established with Muse Spark, with training conducted at Meta’s Prometheus data center. The company intends to integrate Muse Wonder into more features within its Muse platform, initially targeting subscribers. The October rollout provides developers with immediate access to Muse Wonder, while reserving the larger Muse Wonder Full model for Muse Code subscribers.