Meta introduces two Muse Wonder models in push against AI rivals
Muse Wonder will soon power Meta’s personal AI agent for subscribers, while Muse Wonder Full is available through its coding tool.
Meta launches Muse Wonder and Wonder Full in frontier AI push
Wonder will be offered through Meta’s API, while Wonder Full will be restricted to Muse Code subscribers. Both models will power the Muse agent.
Meta’s Superintelligence Lab introduced Muse Wonder and Muse Wonder Full on October 12, 2026, marking the company’s first attempt at frontier AI models. Meta said the larger models offer stronger performance on demanding tasks, including coding, than its lightweight Muse Spark model.
For Wonder, Meta Model API pricing is $4 for each million input tokens and $20 for each million output tokens. Wonder Full will not have API access; availability will be limited to Muse Code subscribers. Meta also plans to use both models in its Muse agent.
Tests published by Meta in its model card put the Wonder models at benchmark levels comparable to Fable 5.1 and GPT-6 Astra. The company emphasized their capacity to remember and combine information supplied by users over extended periods, describing that capability as “active memory.” Meta’s testing also placed Muse Wonder ahead of Fable, Opus and Astra on MemoryBench, which evaluates continual learning and memory construction frameworks.
Research scientist Kavya Kopparapu said the development effort focused heavily on helping the models remember information important to users over days, weeks and months. Meta said training took place at its Prometheus data center in central Ohio.