Meta released its first frontier models, Muse Wonder and Muse Wonder Full. Wonder is larger and more capable than Muse Spark. It scores higher on benchmarks and complex tasks like coding. Wonder matches Fable 5.1 and GPT-6 Astra on benchmarks. Wonder costs $4 per million input tokens and $20 per million output tokens. Wonder Full is only available to Muse Code subscribers. Both models power the Muse agent. They emphasize long-term memory for user information. Wonder scores highest on the MemoryBench benchmark. The models were trained at Meta’s Prometheus data center in Ohio.

• $4 per million input tokens
• $20 per million output tokens
• higher MemoryBench scores than Fable, Opus, Astra