Meta launches Muse Wonder, its largest and most advanced AI model series, to compete with Anthropic and OpenAI

Note to readers: the Business Insider article download is only a partial preview, as the story is reserved for subscribers. This summary is limited to what appears in the accessible text.

Meta has unveiled a new series of AI models called Muse Wonder, which the company says allows it to catch up with labs such as Anthropic and OpenAI. CEO Mark Zuckerberg said on X on Monday that Muse Wonder is "a much larger and more advanced class of models" that puts Meta at the technological frontier.

The series comes in two versions: Muse Wonder, a smaller, faster model that Meta expects people to use for everyday work, and Wonder Full, a larger and more capable version, according to a company statement.

Meta says it built these models to complete complex, multi-step tasks, such as disputing an insurance claim. To that end, they include features such as memory and the ability to try alternative approaches.

The article notes that the launch follows that of Muse, Meta's personal AI assistant, and presents it as a sign that the company's big bet on AI, which raised doubts among investors, is paying off. Meta is spending hundreds of billions of dollars on AI, which prompted some skepticism earlier in the year.

According to a person familiar with the matter, the series was internally code-named Watermelon. In July, Business Insider reported that Meta's head of AI, Alexandr Wang, told staff that Watermelon had caught up with OpenAI and used an order of magnitude more compute than Avocado, the code name for Muse Spark, an earlier family of models. Ahead of the launch, Zuckerberg and other Meta employees teased the release with watermelon-themed posts.

The claim that Meta has reached the frontier should be treated with caution: it comes from the company itself, and the accessible material includes no benchmark results or technical details to back it up.