Meta says its new AI models on par with OpenAI, Anthropic

Meta has unveiled two new advanced AI models, named Wonder and Wonder Full, that the company claims rival the capabilities of industry leaders OpenAI and Anthropic. In a statement on X, CEO Mark Zuckerberg described Wonder as "our largest and most capable model yet" and stated it is "rivals the big (boys)" referring to major AI labs.

Meta's CEO recently pushed the company to compete with top AI labs, especially OpenAI and Anthropic, as it seeks to diversify revenue streams beyond advertising. In June 2025, Meta invested $14 billion in Scale AI and hired its co-founder Alexandr Wang as the company's chief AI officer to oversee its Superintelligence Labs division.

The new models are set to be integrated into Meta's Muse personal agent app for paid subscribers. Benchmarking data from Artificial Analysis suggests Wonder outperforms Meta's previous models and various models from Anthropic, OpenAI, and Google. However, Meta's new AI models are among the most expensive. Despite concerns about safety, Meta has committed billions towards AI investments and delayed the initial launch of Muse to focus on safety and security.

In August, Meta introduced Muse Code, a product competing with automated coding tools from OpenAI and Anthropic, followed by Muse, an agentic app capable of performing tasks autonomously. Meta also unveiled a handheld AI hardware product, Muse Charm, enabling users to interact with their Muse agent outside of a smartphone. The industry is still grappling with the challenge of creating AI-powered hardware.