Meta announced Muse Wonder and Muse Wonder Full on October 12, presenting them as new frontier models aimed at personal and professional agent workflows. Meta says Muse Wonder is available through its Model API and Muse Code, while Muse Wonder Full is available to Muse Code subscribers.
A model launch tied to agent memory
The company describes both models as supporting agentic tasks and highlights Active Memory and Focus features intended to help agents retain context and concentrate on longer assignments. The announcement describes product scope and access, but it does not by itself establish independent performance across real-world agent tasks.
Meta’s own research and newsroom pages provide the release details. No independent benchmark reproduction was located for this edition, so capability language remains attributed to the company.