Meta ships two Muse Wonder models as Zuckerberg rejects AI slowdown
The company is counting on subscription revenue from its Muse assistant to justify $145 billion in planned compute spending in 2026.
By The Carnegie Standard Staff
· 1 min read

Meta chief executive Mark Zuckerberg said the company has reached the AI "frontier" with two models released in the week of October 13, Muse Wonder and Muse Wonder Full, in an interview with the Financial Times reported by Fortune on October 13.
The models follow Meta's Muse personal assistant, released in September 2026, which climbed to the top of Apple's App Store and prompted OpenAI to release a competing product called Dots. Meta's own benchmarks place the Muse models ahead of OpenAI's GPT-6 and Anthropic's Claude Fable 5.1 on coding and memory tasks. Two outside evaluators, Artificial Analysis and Vals, also rated the models highly, with Vals ranking Wonder first on cybersecurity tasks.
The new models are intended to support paid tiers of the Muse assistant, which costs $20 or $100 a month above a free version. Investors have grown more positive on that revenue stream, Reuters reported, a factor in supporting Meta's planned $145 billion of compute spending in 2026, double its 2025 level.
Zuckerberg again dismissed calls from OpenAI, Anthropic and others for a coordinated industry slowdown, arguing Meta already builds in safety work. Bill Nguyen, cofounder of Olive.is, told Fortune of Meta's progress: "They caught up way faster than I thought."
This story was first reported by Fortune.




Comments