MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has unveiled Muse Wonder and Muse Wonder Full, its most advanced artificial intelligence systems to date. Designed for tasks involving programming, extended AI agents, and intricate personal or professional projects, these models aim to assist users in navigating insurance claims, understanding leases, analyzing financial data, and developing functional applications. Muse Wonder became accessible on October 12 via the Meta Model API and Muse Code, while Muse Wonder Full was released exclusively for Muse Code subscribers.
This launch marks Meta’s expansion into AI systems capable of managing multi-step workflows, moving beyond simple prompt-response interactions. Muse Wonder can organize longer tasks, assign subtasks to subordinate agents, utilize software tools, and recover from errors during execution. Additionally, Meta engineered these models for use across various platforms, including browsers, desktop applications, and mobile devices. Both models support a context window nearing one million tokens, enabling them to process extensive data within a single session.
To support prolonged and customized tasks, Meta incorporated two features: Active Memory and Active Focus. Active Memory retains relevant information across conversations and retrieves pertinent details when needed for new tasks. Active Focus assists the models in maintaining progress during challenging tasks, exploring alternative approaches, and recovering after errors. Meta also states that the models can seek clarification from users and verify actions before executing critical operations.
Models focus on programming and persistent AI agent functions
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, based on the company’s developer details. Muse Wonder Full is only available via paid Muse Code subscriptions, not through separate API pricing. For developers with higher coding demands, Meta offers Power Usage at $50 a month and Maximum Usage at $200 a month.
Meta reports that Muse Wonder Full achieves leading results in several agentic and coding benchmarks included in the initial release. Tests like DeepSWE and SWE-Marathon, which evaluate software engineering and long-running coding efficiency, are among the benchmarks highlighted. 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 surpass the earlier Muse Spark generation according to the evaluator’s index.
Enhanced safety features accompany expanded model capabilities
Meta trained the Muse Wonder family to prevent sharing private data without consent and to decline harmful requests. The models also include cybersecurity functionalities that help developers identify vulnerabilities in their code and improve their defensive measures. The company states it assesses its most advanced models under the Meta Superintelligence Scaling Framework and plans to publish a comprehensive report on additional safety evaluations following the initial launch.
Meta began a major overhaul of its AI initiatives approximately 14 months prior to releasing Muse Wonder. The company explains that these models build upon the foundations laid by Muse Spark, with training conducted at its Prometheus data center. The rollout of Muse Wonder aims to integrate the models into more features within its Muse product, starting with subscribers. The October launch provides developers immediate access to Muse Wonder, while reserving the larger Muse Wonder Full for Muse Code subscribers.