Meta Unveils Muse Wonder AI Models, Claims Frontier Capabilities

Meta announced two new artificial intelligence models, Muse Wonder and Muse Wonder Full, according to wired.com. The company says the models will be available through an API and through Muse Code, a tool for writing software with AI.

Meta describes Muse Wonder Full as having frontier capabilities, ranking among the most capable AI models in the world. The company also says the best version of the model will power features across its platforms, including its Muse personal agent.

According to Meta, the largest version of Muse Wonder outperforms Anthropic's Claude Fable 5.1 and ranks just below Claude Opus 5.5 on the Artificial Analysis Intelligence Index. Meta also says Muse Wonder outperforms OpenAI's GPT-6 Astra, Claude Fable 5.1 and Claude Opus 5.5 on some specialized benchmarks, including legal document work and software building.

Muse Wonder includes two new features, Active Memory and Active Focus, designed to keep the model locked into a stated goal and prevent confusion. Meta says it trained Muse Wonder to refuse to generate dangerous outputs such as bioweapon recipes or hacking help.

Meta did not disclose details of how the Muse Wonder models were built or trained. The company also did not disclose whether agents powered by Muse Wonder misbehaved in testing or what it did to mitigate that issue.

Background

Last year, Mark Zuckerberg rebooted Meta's artificial intelligence efforts by sinking billions into the newly formed Meta Superintelligence Labs and hiring former Scale AI CEO Alexandr Wang to lead it, according to wired.com. Meta was once a trailblazer in AI, and its Llama series were the first frontier open-weight offerings from a big lab. The company fell behind after delivering several disappointing versions of the Llama model, as Anthropic, OpenAI and Google rapidly introduced more capable models. In April of this year, Meta announced the first model from its new lab, Muse Spark.