Meta has launched Muse Wonder and Muse Wonder Full, its most capable frontier models designed specifically for personal and professional AI agents. Built to handle long-running agent work such as coding, legal, and finance tasks alongside computer use, the models feature a one-million-token context window supporting text, image, video, and audio inputs. Muse Wonder is available now via the Meta Model API and Muse Code, while Muse Wonder Full is restricted to paid Muse Code plans starting at fifty dollars per month.
Trained at Meta's Prometheus data center, the models introduce two key behaviors named Active Memory and Active Focus, allowing them to carry user context across conversations and operate browsers as well as desktop and mobile applications. In Meta's benchmark evaluations, Muse Wonder Full leads six of fourteen rows, including strong showings on DeepSWE and SWE-Marathon, though competing models like Claude Opus 5.5 still lead in categories such as ProgramBench and OSWorld. Meanwhile, Muse Wonder achieved a top score of 85.6 on the internal MemoryBench evaluation.
Developers can access the models through the Meta Model API using an OpenAI-compatible endpoint, with pricing set at four dollars per million input tokens and twenty dollars per million output tokens. The release aims to strengthen Meta's position in autonomous agent infrastructure by providing advanced reasoning capabilities for complex software and administrative workflows.