Muse Wonder is available through the Meta Model API under the identifier muse-wonder-1.0 and in the Muse Code coding assistant. Muse Wonder Full, the stronger variant, is limited to Muse Code subscribers and has no API access. Meta says Wonder will soon power the Muse app, starting with subscribers. Both models were trained at the Prometheus gigawatt-scale data center, and Meta claims state-of-the-art scores on Harvey's legal benchmark, FocusBench and SWE-Marathon. Its pitch rests on Active Memory, which carries context across conversations, and Active Focus, which keeps an agent working through blockers. API pricing is $4 input and $20 output per million tokens, $0.20 for cached input, plus $2.50 per 1,000 web searches, according to VentureBeat. Artificial Analysis estimates an average index task costs $4.52 on Wonder, against $5.98 on Opus 5.5 and $3.26 on GPT-6 Astra. For builders, this is Meta's first frontier-tier offer for agent developers, but the strongest model sits inside Meta's own coding product. Teams comparing vendors should test Wonder on their own long-running tasks, since per-token price and per-task cost diverge.