Meta releases Muse Wonder models to challenge OpenAI and Anthropic
Meta launched its first family of frontier models, claiming performance parity with top competitors in agentic tasks. The release marks a direct competitive entry into the high-end AI market, leveraging Meta's massive user base.
Meta has released Muse Wonder, its first family of frontier models designed for personal and professional agentic work. The lineup includes a top-tier variant, Muse Wonderful Max, and a lower-tier version. On the Personal Super Intelligence Index, the top-tier model scored 59, while Artificial Analysis ranked it second on its Intelligence Index with a score of 56, trailing Opus 5-5 by two points. The model achieved a 79% score on the Deep SWE 1.1 benchmark, leading GPT-6 Astra and Opus 5.5 by nearly five points. However, community feedback noted significant underperformance in 3D tasks and video game generation, a gap Alexander Wang acknowledged by stating the team did not explicitly train on 3D demo games. Meta is positioning these models to leverage its existing data infrastructure, noting that 3.6 billion people use a Meta app daily. The top-tier performance is available to subscribers of Muse Code, Meta's coding platform, while the base Muse app remains free. Pricing for the Wonderful variant is set at $4.52 per task, a threefold increase from the previous MuseSpark 1.3 model. Meta stock rose 4% on the news. Mark Zuckerberg stated in a Financial Times interview that Meta has reached the frontier and model development will accelerate, rejecting the pacing approach adopted by other labs. Dan Ives of Yorkville maintained an Outperform rating with a price target of $900, arguing that Meta's distribution advantage makes the new models a significant competitive threat to specialized AI labs. Meta is investing between $130 billion and $145 billion this year.