Meta says new muse model hits AI frontier after aggressive push

Meta Platforms Inc. unveiled a new artificial intelligence model called Muse Wonder on Monday, marking a significant milestone in the company's multibillion-dollar push to rebuild its AI organization. CEO Mark Zuckerberg's aggressive hiring efforts appear to have paid off, as Muse Wonder boasts "frontier capabilities" that Meta claims put it on par with OpenAI's ChatGPT and Anthropic PBC's Claude in coding and handling financial or legal inquiries.

The model's release comes after a year and a half of efforts from Zuckerberg, who committed significant resources to AI development and recruited Alexandr Wang to lead the new AI unit. Wang, Meta AI Chief, described Muse Wonder as "true frontier models" and a "big moment" for the company. The new model aims to differentiate Meta from competitors by focusing on broad consumer uses, such as a personal AI agent called Muse, which will power the subscription version of the app.

Muse Wonder can retain information across days, weeks, and months, allowing users to interact with the AI agent without re-explaining themselves. The model's active memory and focus capabilities will enable it to tackle more complex, ambiguous tasks. Wang and his Superintelligence Labs team are already working on even larger and more efficient models, alongside efforts to reduce the cost of serving the technology to billions of users.

The new model is available to developers through the Meta Model API and Muse Code, and Wang encourages skeptics to try it and judge its performance for themselves.