Meta has released Muse Wonder and Muse Wonder Full, its first frontier AI models, putting the company back in the race with OpenAI, Anthropic and Google after more than a year of rebuilding its artificial intelligence efforts. The models were announced on October 12, 2026, and arrive as a direct sequel to Muse Spark, the lightweight model that launched Meta Superintelligence Labs earlier this year.
The launch was accompanied by a candid Q&A with CEO Mark Zuckerberg and Chief AI Officer Alexandr Wang on the Cheeky Pint podcast, where the pair discussed committing $100 billion or more to AI infrastructure before knowing whether the models would work, and laid out their vision for agentic commerce and personal AI agents that belong to billions of people.
Muse Wonder specs and features
According to Meta's own testing, Muse Wonder and Muse Wonder Full reach benchmark scores comparable to Fable 5.1 and GPT-6 Astra. Wired reports that the largest version outperforms Anthropic's Claude Fable 5.1 and ranks just below Claude Opus 5.5 on the Artificial Analysis Intelligence Index, while Meta claims it beats OpenAI's best model on specialized tasks such as legal document work and software development. Benchmarks are not always a reliable measure of real-world performance, but the results mark a significant step forward for Meta's models.
Meta says the models are especially strong at legal, financial and health-related tasks and deliver frontier coding performance. Both were trained at Meta's Prometheus data center in central Ohio, which came online earlier this year.
- Active Memory: carries what matters to you across conversations and recalls relevant details for the task at hand
- Active Focus: keeps an agent working toward your goal when a task gets hard



