Meta unveiled two new "frontier" AI models called Wonder and Wonder Full, claiming they perform on par with OpenAI's ChatGPT and Anthropic's Claude.
"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," chief executive Mark Zuckerberg wrote on X.
Chief AI officer Alexandr Wang, who joined Meta as part of its $14 billion investment in Scale AI last year, said Wonder "rivals the big (boys)" in his own post on the platform.
Benchmarked claims
Meta cited benchmark data from Artificial Analysis showing Wonder outperforming its predecessor Spark, launched in April as a break from the Llama open-source line, as well as rival models from OpenAI, Anthropic and Google.
Artificial Analysis's figures also place Wonder among the most expensive models on the market.
The new models will reach paid subscribers of Muse, Meta's agentic assistant app, following the August launch of a coding-focused product, Muse Code, and a handheld companion device, Muse Charm, shown at the Connect conference last month.
Spending boom
Meta's AI spending could reach $145 billion this year, part of a buildout the company has pursued despite signing a "morally binding" safety pledge with rivals including Google, OpenAI, Anthropic, xAI and Nvidia in September.
Zuckerberg has said Meta delayed Muse for months to work on safety, even as Anthropic chief executive Dario Amodei told the United Nations last month that labs "will slow down as much as necessary" to keep new releases safe.
