MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has launched its latest and most powerful artificial intelligence models, Muse Wonder and Muse Wonder Full. These systems are specifically engineered for coding, managing long-duration AI agents, and executing complex personal or professional tasks. According to Meta, users can leverage these models to handle insurance claims, comprehend lease agreements, analyze financial data, and develop functional applications. Muse Wonder was made available on October 12 via the Meta Model API and Muse Code, with Muse Wonder Full debuting exclusively for Muse Code subscribers.
This launch marks Meta’s expansion into AI systems capable of multi-step completion rather than merely responding to individual prompts. Muse Wonder is designed to plan extended tasks, assign work to sub-agents, utilize software tools, and recover from errors during operation. Additionally, Meta developed these models for use across various platforms, including browsers, desktop programs, and mobile applications. Both models boast a context window of approximately one million tokens, enabling them to process extensive information within a single session.
To support more personalized and prolonged tasks, Meta introduced features named Active Memory and Active Focus. Active Memory maintains relevant information throughout conversations and retrieves details when they are pertinent to a new task. Active Focus assists models in continuing work when facing difficulties, testing alternative approaches, and recovering after errors. Meta also states that the models can ask users for clarification and verify actions before executing significant operations.
Models designed for coding and persistent AI agent tasks
Through the Meta Model API, Muse Wonder is priced at $4 per million input tokens and $20 per million output tokens. Cached input costs are $0.20 per million tokens, as specified in the company’s developer documentation. Muse Wonder Full is available only via paid Muse Code subscription plans, not as a separate API offering. Developers with higher coding demands can opt for Power Usage at $50 a month and Maximum Usage at $200 a month.
Meta reports that Muse Wonder Full surpasses previous models in multiple agentic and coding benchmarks, with results published alongside the release. Among the assessments used to evaluate software engineering and long-duration coding performance are DeepSWE and SWE-Marathon. Independent evaluator Artificial Analysis assigned Muse Wonder an Intelligence Index score of 52, and Muse Wonder Full received a score of 56 during maximum reasoning efforts. These scores position the models above Meta’s earlier Muse Spark generation on the evaluator’s index.
Enhanced safety features accompany advanced capabilities
Meta has trained the Muse Wonder series to refrain from sharing private data without permission and to decline harmful requests. The models also retain cybersecurity functionalities that enable developers to identify vulnerabilities in their code and bolster security defenses. The company states that its most capable models are evaluated under the Meta Superintelligence Scaling Framework, and plans to publish a comprehensive report on additional safety assessments following the initial launch.
Meta initiated a major overhaul of its AI initiatives approximately 14 months prior to launching Muse Wonder. The new models build upon the foundation laid by Muse Spark, with training carried out at the Prometheus data center. The company intends to integrate Muse Wonder into more experiences within its Muse platform, beginning with subscribers. The October rollout provides developers immediate access to Muse Wonder, while reserving the larger Muse Wonder Full for Muse Code subscribers.