Meta Debuts First Frontier Models Muse Wonder and Wonder Full
Meta's Superintelligence Lab releases a pair of state-of-the-art models on Monday, Muse Wonder and Muse Wonder Full, marking the company's first frontier model release. The larger Wonder models outperform the lightweight Muse Spark, delivering higher benchmark scores and stronger results on complex tasks like coding. Meta reports that Wonder and Wonder Full achieve benchmark results comparable to Fable 5.1 and GPT-6 Astra, calling the release "a major leap toward personal superintelligence."
Pricing for the models places Meta firmly in competition with rivals. Wonder, available through the Meta Model API, costs $4 per million input tokens and $20 per million output tokens, roughly matching GPT-6 Sol and Opus 5.5. Wonder Full, however, is reserved exclusively for Muse Code subscribers and is not served through the API. Both models also power the Muse agent.
A standout feature is "active memory," Meta's approach to helping models retain and synthesize user information over extended periods. The company highlights MemoryBench, a benchmark testing continual learning and memory construction, where Muse Wonder now scores higher than Fable, Opus, or Astra. The release follows Alexandr Wang's restructuring of Meta's AI efforts last August, and the models were trained at Meta's Prometheus data center in central Ohio, which came online earlier this year.