MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has launched Muse Wonder and Muse Wonder Full, its most sophisticated artificial intelligence models yet. These new systems are tailored for coding, extended AI agents, and handling complex personal or professional tasks. Meta claims the models assist users with insurance claims, lease comprehension, financial analysis, and application development. Muse Wonder debuted on October 12 via the Meta Model API and Muse Code, with Muse Wonder Full available exclusively to Muse Code subscribers.
This rollout marks Meta’s expansion into AI capable of multi-step processes, moving beyond simple prompt responses. Muse Wonder can plan multi-stage tasks, allocate work to sub-agents, utilize software tools, and recover from errors. The models are designed for advanced use across browsers, desktop software, and mobile apps. Both support a context window of roughly one million tokens, enabling the processing of extensive data in a single session.
Meta introduced features called Active Memory and Active Focus to facilitate longer, more personalized interactions. Active Memory retains relevant information across conversations and retrieves details as needed, while Active Focus helps the models maintain task continuity when challenges arise, test alternatives, and recover from mistakes. The models can also seek user clarification and verify actions before execution.
Targeting coding and persistent AI agents
Pricing for Muse Wonder is $4 per million input tokens and $20 per million output tokens via the Meta Model API. Cached input costs $0.20 per million tokens, according to developer data. Muse Wonder Full is accessible only through paid Muse Code plans, not separate API options. Developers with larger workloads can opt for Power Usage at $50 monthly and Maximum Usage at $200 monthly.
Meta reports that Muse Wonder Full surpasses previous models in several agentic and coding benchmarks, including DeepSWE and SWE-Marathon tests for software engineering and long-term coding. 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, outperforming Meta’s earlier Muse Spark models.
Enhanced safety measures accompany expanded capabilities
Meta trained the Muse Wonder series to avoid sharing private data without permission and to decline harmful requests. The models also incorporate cybersecurity functions that assist developers in identifying flaws and bolstering defenses. These models are evaluated under the Meta Superintelligence Scaling Framework, with plans to publish a comprehensive safety report following the initial release, including safety assessments.
Meta began revamping its AI initiatives roughly 14 months prior to launching Muse Wonder, building on the foundation laid by Muse Spark at its Prometheus data center. The company intends to expand Muse Wonder’s integration into its Muse platform, starting with subscribers. The October launch provides immediate access to Muse Wonder for developers, reserving Muse Wonder Full for Muse Code members.