MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has introduced Muse Wonder and Muse Wonder Full, an innovative lineup of sophisticated artificial intelligence models. Designed for software development, AI agents, and intricate daily responsibilities, these systems can support functions such as lease management, insurance claim processing, financial data analysis, and application creation. Meta launched Muse Wonder via its Model API and Muse Code on October 12, with Muse Wonder Full accessible to users holding eligible Muse Code subscriptions.
Focusing on multi-step tasks rather than single-response queries, the models enable planning, digital tool integration, and the subdivision of large projects among specialized sub-agents. They are capable of functioning across browsers, desktop software, and mobile platforms. Meta engineered the system to recover seamlessly if a selected approach fails. With context windows supporting approximately one million tokens, they can handle sizable documents, codebases, and extensive materials within a single session.
In addition, Meta introduced Active Memory and Active Focus alongside the Wonder models. Active Memory retains essential details from prior conversations and recalls them as necessary. Active Focus assists AI agents in tackling complex issues by maintaining progress and adjusting strategies after errors. The models can ask for clarification when instructions lack critical information and seek confirmation before executing sensitive or significant actions. These features aim to extend longer work sessions without the need for users to restart tasks repeatedly.
Meta’s new models broaden agent and coding functionalities
Pricing for Muse Wonder is set at $4 per million input tokens via the Meta Model API. Output costs are $20 per million tokens, while cached input tokens are priced at $0.20 per million. Instead of separate API charges, Meta offers Muse Wonder Full through paid Muse Code plans, including a Power Usage plan at $50 per month and a Maximum Usage plan costing $200 monthly, tailored for developers with intensive coding needs.
Meta shared benchmark results evaluating software engineering, coding agents, and long-duration technical tasks. Key tests included DeepSWE and SWE-Marathon, with Artificial Analysis assessing both models at maximum reasoning effort. Muse Wonder received an Intelligence Index score of 52, while Muse Wonder Full achieved a score of 56. These results surpass Meta’s earlier Muse Spark models on the evaluator’s scale, indicating a clear improvement in reasoning performance.
Safety and privacy safeguards accompany the debut
Meta emphasizes that the Wonder models incorporate measures to protect privacy, prevent harmful requests, and ensure cybersecurity compliance. They are trained to withhold private information unless user permission is granted and to decline requests involving harmful activities while maintaining legitimate security functions. Developers can leverage these models to identify vulnerabilities in their software and enhance defenses. Meta assesses its most advanced AI systems under its Superintelligence Scaling Framework and plans to publish additional details about safety testing for the new models.
This launch comes after a significant restructuring of Meta’s AI efforts that began approximately 14 months earlier. The company states that Muse Wonder builds upon the technical groundwork laid by Muse Spark. Development utilized computing resources at Meta’s Prometheus data center. The firm also intends to incorporate Muse Wonder technology into other features within the Muse platform, starting with subscribers. Currently, developers can access Muse Wonder, while Muse Wonder Full remains available through qualifying Muse Code subscription plans.