Meta 發佈 Muse Wonder 系列 AI 模型,稱其為歷來最大、最強模型
Meta says it's launching the biggest and best AI models it has ever built
Meta 發佈 Muse Wonder 系列 AI 模型,並稱其已追上 Anthropic 和 OpenAI 等實驗室。系列包括面向日常工作的較小、較快 Muse Wonder,以及規模更大、能力更強的 Wonder Full;Meta 表示,這些模型可處理複雜多步任務,並具備記憶和嘗試替代方案的功能。
Meta says it's launching the biggest and best AI models it has ever built
You're currently following this author! Click to unsubscribe from email alerts.
Meta has launched a new series of AI models, Muse Wonder, that it says has caught up to labs like Anthropic and OpenAI.
"Muse Wonder is a much larger and more advanced class of models that brings us to the frontier," CEO Mark Zuckerberg posted on X on Monday.
It’s the latest sign that Meta’s big bet on AI — which has raised questions from investors — is paying off after the launch of its personal AI assistant, Muse. Meta is spending hundreds of billions of dollars on AI, prompting some investor doubts earlier this year.
The Wonder series comes in two versions: Muse Wonder, a smaller, fast model Meta expects people to use for daily work, and Wonder Full, a larger and more capable version, the company said in a statement.
Meta built its new AI models to complete complex multi-step tasks, such as disputing an insurance claim. To handle that, the AIs have features such as memory and the ability to try alternative approaches.
The Muse Wonder series was internally codenamed Watermelon, according to a person familiar with the matter.
In July, Business Insider reported that Meta AI chief Alexandr Wang had told staff that Watermelon had caught up to OpenAI.
"Watermelon uses an order of magnitude more compute than Avocado," he said at the time, referring to Meta's codename for Muse Spark, a prior family of models.
In the lead-up to Muse Wonder's launch, Zuckerberg and other Meta employees teased the release with watermelon-themed posts:
Have a tip? Contact this reporter securely on Signal at charlesrollet.12 or +1-628-282-2811. Use a personal email address and a nonwork device; here's our guide to sharing information securely.
Read next
You're currently following this author! Click to unsubscribe from email alerts.
Charles Rollet is Business Insider's tech correspondent in San Francisco, where he focuses on breaking major news about the AI industry. In 2025, he led an investigation which revealed that Scale AI had inadvertently exposed confidential data about its workers and clients. In 2026, he was the first to break the news that Meta had halted a controversial keystroke tracking program.Prior to joining BI, Charles worked at TechCrunch covering startups and VC in San Francisco. Before that, he was based in Southeast Asia for a decade, where his work won a Gerald Loeb award and was nominated for a Society of Publishers in Asia award. His reporting on the surveillance industry led to US human rights sanctions on five Chinese security camera manufacturers.Charles is especially interested in how frontier AI models are built and trained. He has a master's degree in business reporting from NYU and a bachelor's degree in history from Northwestern.You can contact Charles securely on Signal at charlesrollet.12 or +1-628-282-2811. Charles keeps sources anonymous. Please use a non-work device to reach out.
來源:Business Insider Tech · businessinsider.com