Meta Launches Two New AI Models, Wonder and Wonder Full, in Challenge to OpenAI and Anthropic
Meta unveiled two new "frontier" artificial intelligence models on Monday, named Wonder and Wonder Full, which the company says perform at a level comparable to OpenAI's ChatGPT and Anthropic's Claude.
The announcement was made by chief executive Mark Zuckerberg in a post on X. "Muse Spark was our first family of models, and now Muse Wonder is a much larger and more advanced class of models that brings us to the frontier," he wrote.
The performance claims are Meta's own. The company cited benchmarking data from Artificial Analysis, an independent firm, which Meta said shows the new Wonder models outperforming its earlier Spark models as well as models developed by Anthropic, OpenAI and Google. Artificial Analysis's own website lists Wonder among the most expensive models on the market.
Wonder is Meta's first new model since April, when the company launched a frontier model called Spark. That launch marked a shift away from Meta's open-source Llama family, which had fallen behind offerings from rival laboratories.
The change reflects a broader shift in strategy. Meta built much of its reputation in AI on Llama, released under open-source licences that allowed outside developers to use and modify the models. Its recent releases, including Spark and now Wonder, follow the frontier-model approach of OpenAI, Anthropic and Google, where access is controlled and, in some cases, sold.
Alexandr Wang, co-founder of the San Francisco-based AI startup Scale AI, posted on X that Wonder is "our largest and most capable model yet". Meta invested $14 billion in Scale AI in June 2025 and brought Wang into the company as part of the deal. He was subsequently appointed chief AI officer, overseeing a newly formed division called Superintelligence Labs.
The new models can be used in Meta's personal agent app, known as Muse, though access is currently limited to paid subscribers.
For Meta, the push into frontier AI is part of a wider effort to compete with leading AI laboratories and to diversify beyond advertising, from which the company has historically drawn most of its revenue. Zuckerberg has pressed Meta to close the gap with OpenAI and Anthropic over the past year or so, as the contest for customers and for leadership in the technology has intensified.
In July, the Facebook and Instagram parent said its spending on AI-related investments could reach $145 billion this year. The company reports more than 3.6 billion daily active users across its platforms.
The launch comes as governments and researchers continue to raise concerns about the safety and security of increasingly powerful AI systems. Meta has committed billions of dollars to the global race to lead in AI even as those concerns persist.
What remains to be seen is how the new models perform in everyday use, as opposed to controlled benchmarks, and whether customers will pay premium prices for them. Independent evaluations and wider user testing over the coming months are likely to shape that assessment.
This report is based on Meta's announcement and public statements by its executives, along with data published by Artificial Analysis.
