Meta's Muse Wonder lands at 59.1, Zuck now says superintelligence is "personal" 📈

Meta released Muse Wonder and Muse Wonder Full on Monday, the first models from its Meta Superintelligence Labs that the company is positioning head-to-head with offerings from OpenAI and Anthropic. "Muse Wonder is a major leap towards delivering personal superintelligence for everyone," Mark Zuckerberg wrote on Threads. "Congrats to everyone on our team for achieving this milestone. I'm looking forward to seeing how you all use it." In a blog post, Meta described the Muse Wonder series as "our most capable models for personal and professional tasks," adding that "together, they mark the next step on our scaling ladder and a major leap toward personal superintelligence."

The two models center on Active Memory, which carries user context from one conversation into the next, and Active Focus, which keeps the system iterating toward a stated goal and checking in before executing major actions. Muse Wonder is available immediately through the Meta Model API and Muse Code, Meta's coding agent, while the larger Muse Wonder Full is restricted to paying Muse Code subscribers. Both hold 1 million tokens in their context window, according to Meta's model card.

The launch follows Muse Spark, which debuted earlier this year as a small, closed model and effectively marked the end of Meta's open-source Llama line. Muse Spark ranked among top applications after release. The shift followed the April 2025 arrival of Llama 4, which arrived with high-profile benchmarks and later drew criticism when independent tests suggested the version posted to a public leaderboard behaved differently from the publicly downloadable build. Zuckerberg responded with a roughly $14 billion deal for Scale AI and the appointment of its founder, Alexandr Wang, to lead Meta Superintelligence Labs.

On Meta's Personal Superintelligence Index, an in-house composite of personal and professional assistant tasks where higher is better and the exact formula has not been published, Muse Wonder Full scored 59.1. Anthropic's Claude Opus 5.5 posted 55.2, OpenAI's GPT-6 Astra reached 54.5, and Anthropic's Claude Fable 5.1 came in at 49.9. On SWE-Marathon, a coding benchmark, Wonder Full scored 57.5 versus 46.3 for Opus 5.5 and 42.5 for Astra.

The release lands weeks after OpenAI shipped GPT-6 Astra on September 3 and Anthropic followed with Claude Opus 5.5 on September 22, which the company says matches its higher-priced Claude Fable 5.1 on most workloads. Meta's announcement also follows Elon Musk's SpaceXAI releasing Grok 4.7 last month, underscoring the continued pace of frontier model competition across the sector.