Meta has launched Muse Wonder and Muse Wonder Full, the company's first frontier-class artificial intelligence models. Developed by the Superintelligence Lab, these systems are designed for complex tasks, including programming, and demonstrate performance metrics comparable to Fable 5.1 and GPT-6 Astra.
The models are available via the Meta Model API. Pricing is set at $4.00 per million input tokens and $20.00 per million output tokens, positioning them competitively with GPT-6 Sol and Opus 5.5. Training for these models took place at Meta’s Prometheus data center in central Ohio.
A key feature of the new models is active memory, which allows the AI to retain and synthesize user information over extended periods. Meta introduced the MemoryBench benchmark to evaluate this capability, reporting that Muse Wonder outperforms Fable, Opus, and Astra in memory-related tasks.
This release marks a shift in Meta's AI strategy following the restructuring of its research efforts by Scale AI founder Alexandr Wang in August 2025. While the company previously focused on lightweight solutions like Muse Spark, Muse Wonder represents its first entry into the state-of-the-art frontier model market.
