Meta unveils Muse Wonder AI to rival OpenAI and Anthropic

Meta unveiled two artificial intelligence models on Monday designed to power personal and professional AI agents, positioning the technology giant to challenge systems developed by OpenAI, Anthropic, and Google in the rapidly expanding market for automated task management.

Advanced capabilities

Muse Wonder and Muse Wonder Full represent the company's most advanced AI systems to date, featuring Active Memory that retains user information across conversations and Active Focus that enables persistent problem-solving when encountering obstacles. The models can execute long-running tasks, navigate complex coding projects, delegate work to smaller AI agents, and recover from errors without abandoning objectives, according to the company's announcement.

The systems are designed to handle specialized professional domains including law, finance, and health. Meta noted they can assist users with understanding leases, navigating insurance claims, analyzing finances, and building software applications.

Market rollout

Muse Wonder is now available to developers through the Meta Model API and the company's Muse Code programming tool, while the more capable Muse Wonder Full requires a paid subscription starting at $50 monthly. Meta said the systems will later integrate into its Muse personal assistant for subscribers, expanding access beyond the initial developer focus.

Meta published benchmark results claiming the models match or outperform some OpenAI and Anthropic systems on selected coding and professional tasks, with independent testing expected to assess broader performance in the coming weeks.

Strategic positioning

Chief Executive Officer Mark Zuckerberg has framed the launch as part of Meta's push toward "personal superintelligence," describing AI systems capable of understanding users' long-term goals and preferences. Unlike earlier Meta models released with downloadable weights, the company did not announce plans to open-source the Wonder family.

The company said it trained the Wonder models to protect private information, refuse assistance with harmful activity, and strengthen defenses against cyberattacks, with a detailed safety report expected to be released later.