Meta has released Muse Wonder and Muse Wonder Full, its first attempt at a frontier model, first reported by TechCrunch. The larger models are positioned as more capable than Muse Spark, with Meta citing higher benchmark scores and stronger results on complex tasks such as coding, as well as an emphasis on retaining and synthesizing user-provided information.
Meta’s testing gives the pair benchmark scores comparable to Fable 5.1 and GPT-6 Astra. The company also says Muse Wonder scores higher on MemoryBench than Fable, Opus or Astra.
Meta calls its focus on retaining and synthesizing information provided by users active memory. Research scientist Kavya Kopparapu said in a launch video that the team worked to help the model remember information important to users over days, weeks, months and beyond.
Meta splits access by version: Muse Wonder costs $4 per million input tokens and $20 per million output tokens through the Meta Model API, roughly equivalent in cost to GPT-6 Sol or Opus 5.5. Muse Wonder Full is available only to Muse Code subscribers and will not be served through the API; both models will power the Muse agent.
The models mark Meta’s first frontier-model release since Scale AI founder Alexandr Wang restructured the company’s AI efforts in August of last year. Since then, Meta has released lightweight Muse Spark and Muse agent while avoiding the kind of state-of-the-art model releases OpenAI and Anthropic are known for.
The models were trained at Meta’s Prometheus data center in central Ohio, which came online earlier this year.