Meta’s Superintelligence Lab has released Muse Wonder and Muse Wonder Full, its first frontier AI models, marking a step up from the lightweight, low-latency Muse Spark. According to Meta’s model card, the larger Wonder models deliver higher benchmark scores and stronger performance on complex tasks such as coding, achieving results comparable to Fable 5.1 and GPT-6 Astra. In its blog post, Meta described the models as the next rung on its scaling ladder and a major leap toward personal superintelligence.
Wonder will be available through the Meta Model API at $4 per million input tokens and $20 per million output tokens, roughly in line with GPT-6 Sol and Opus 5.5. Wonder Full will be reserved for Muse Code subscribers and not offered via the API, while both models will power the Muse agent.
Reflecting Meta’s focus on long-running personal assistants, the company emphasized a capability it calls active memory, the ability to retain and synthesize user information. Meta said Wonder now outscores Fable, Opus and Astra on the MemoryBench benchmark. Research scientist Kavya Kopparapu said the team worked extensively to ensure the model remembers what matters to users over days, weeks and months.
The release is Meta’s first frontier model since Alexandr Wang restructured its AI efforts last August, and the models were trained at its Prometheus data center in Ohio.
