Meta’s October 12 release of Muse Wonder and the more capable Muse Wonder Full variant, trained at its Prometheus data center with a 1M-token context window and new Active Memory plus Active Focus features, sets the baseline for the next 1.1+ update. The models target long-running agentic tasks in coding, legal, finance, and personal workflows, with Muse Wonder available immediately via the Meta Model API at $4/$20 per million tokens and Full restricted to Muse Code subscribers. Benchmarks show competitive or leading results on internal agent and memory tests while trailing Claude Opus 5.5 on several independent indices. Traders are watching the planned rollout to the Muse app, subscriber feedback on quotas and reliability, and any signals of accelerated scaling or open-weight plans that could accelerate a 1.1 iteration ahead of rivals.