Meta launches Muse Wonder at $4/$20 per million tokens -- but its stronger sibling, Wonder Full, is subscription-only
Meta's first frontier-class release since its AI overhaul ships in two tiers: Muse Wonder on the Meta Model API, and Muse Wonder Full only inside Muse Code's paid plans. The benchmark comparisons are Meta's own.
Meta released Muse Wonder and Muse Wonder Full on October 12, calling them its most capable models and pitching them at personal and professional agent work. Muse Wonder is available now through the Meta Model API and Muse Code, Meta's coding agent, at $4 per million input tokens, $20 per million output tokens and $0.20 per million cached input tokens, with a 1 million token context window. It uses the model id muse-wonder-1.0 and works with existing OpenAI-compatible clients by changing the base URL. Muse Wonder Full isn't served through the API at all: it's available only on Muse Code's Power Usage plan and a new Maximum Usage plan, which Meta says carries 80 times the usage limit of its Everyday Usage plan.
What Meta is claiming
Meta says both models were trained at its gigawatt-scale Prometheus data center and lead on benchmarks for personal and professional agents, memory and coding. Its developer post says Wonder Full tops the DeepSWE and SWE-Marathon coding leaderboards "in independent third-party runs," and that the cheaper Wonder brings performance comparable to Full. TechCrunch, reading Meta's model card, puts the two in the same range as Claude Fable 5.1 and GPT-6 Astra, and notes the API price is roughly what GPT-6 Sol or Claude Opus 5.5 cost. The models' headline features are what Meta calls Active Memory, carrying context about a user across conversations, and Active Focus, sustained work toward a goal that includes checking in before important actions. Meta says a full report with additional safety evaluations "will be available soon."
Why it matters
Muse Wonder adds another frontier-priced option to the API menu, which is useful leverage in any vendor negotiation. Treat the rankings as Meta's own until your team has run the model on its own workloads. The more important detail for anyone tracking spend is the split: the best model in this release can't be metered per token. If developers move to Wonder Full, the cost arrives as Muse Code seat subscriptions, so it won't show up in your API usage. Usage tracking that only reads API keys will miss it.