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

Meta, the social media giant, unveiled two state-of-the-art artificial intelligence models, named Wonder and Wonder Full, on Monday. The company claims these new models possess capabilities comparable to those of OpenAI's ChatGPT and Anthropic's Claude. CEO Mark Zuckerberg stated on X that the new models, part of Meta's Muse Spark family, have ushered the company into a new frontier of AI development.

Zuckerberg's push to compete with leading AI labs like OpenAI and Anthropic has been ongoing for the past year. In June 2025, Meta invested $14 billion in Scale AI, an AI startup, and hired its co-founder Alexandr Wang as its Chief AI Officer. Wang oversees the newly formed Superintelligence Labs division.

Wonder is Meta's first new model since the launch of the frontier model Spark in April. Unlike the open-source model Llama, Wonder and its larger counterpart, Wonder Full, are Meta's most advanced offerings to date. The new models can be used in Meta's paid subscription personal agent app, Muse.

According to Wang, "Wonder is our largest and most capable model yet and rivals the big (boys)," referring to OpenAI and Anthropic. Meta's blog post cited benchmarking data from Artificial Analysis, an independent firm, which suggests that Wonder outperforms its previous models, Spark, as well as models from Anthropic, OpenAI, and Google.

However, the new models are expensive, and Artificial Analysis notes that Wonder is among the most costly. Despite growing concerns about safety and security, Meta has committed billions towards AI-related investments, with Zuckerberg revealing that the company's spending could reach $145 billion this year. Meta boasts over 3.6 billion daily active users across its platforms, including Facebook and Instagram.