Meta releases first frontier models Muse Wonder and Muse Wonder Full on October 12, internal codename Watermelon

Meta Superintelligence Lab released its first frontier models, Muse Wonder and the larger Muse Wonder Full, on October 12. The launch is the first true frontier-class release since Scale AI founder Alexandr Wang restructured Meta's AI efforts last August. The models are larger and more capable than the lightweight Muse Spark and post benchmark scores comparable to Fable 5.1 and GPT-6 Astra. Muse Wonder is served through the Meta Model API at 4 dollars per million input tokens and 20 dollars per million output tokens, priced near GPT-6 Sol and Opus 5.5. The more powerful Muse Wonder Full is not served through the API and is restricted to Muse Code paid subscribers. The models put particular emphasis on active memory, scoring higher than Fable, Opus, and Astra on the MemoryBench benchmark that tests continual learning and memory construction. Research scientist Kavya Kopparapu said in the launch video that the team put tremendous effort into making sure the model remembers what matters to the user over days, weeks, and months. Training was completed at the Prometheus data center in central Ohio that came online earlier this year.