Meta says it's launching the biggest and best AI models it has ever built

Meta launched its Muse Wonder series of AI models on Monday, with CEO Mark Zuckerberg posting on X that "Muse Wonder is a much larger and more advanced class of models that brings us to the frontier," claiming parity with Anthropic and OpenAI. The series ships in two versions — Muse Wonder, a smaller, faster model for daily work, and Wonder Full, a larger, more capable model — both built for complex multi-step tasks such as disputing an insurance claim, with memory and the ability to try alternative approaches. Meta AI chief Alexandr Wang had told staff in July that the internally codenamed Watermelon model caught up to OpenAI and "uses an order of magnitude more compute than Avocado," Meta's codename for the prior Muse Spark family.

Meta says it's launching the biggest and best AI models it has ever built Business Insider https://www.businessinsider.com Meta unveils new AI models, Muse Wonder and Wonder Full, designed for complex tasks. It says it has caught up to rivals like OpenAI and Anthropic. - Meta says its latest AI models have finally caught up with its rivals. - Meta tailored its models, Muse Wonder and Wonder Full, for work and completing multi-step tasks. - Meta CEO Mark Zuckerberg teased the release by posting a photo of a watermelon. Meta has launched a new series of AI models, Muse Wonder, that it says has caught up to labs like Anthropic https://www.machinebrief.com/glossary/anthropic and OpenAI https://www.machinebrief.com/glossary/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 https://www.businessinsider.com/meta-ai-success-nat-friedman-muse-launch-2026-9 . Meta is spending hundreds of billions of dollars on AI, prompting https://www.machinebrief.com/glossary/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 https://www.businessinsider.com/meta-ai-model-catches-up-openai-gpt-5-says-2026-7 had told staff that Watermelon had caught up to OpenAI. "Watermelon uses an order of magnitude more compute https://www.machinebrief.com/glossary/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: Muse ordered me 🍉 how thoughtful pic.twitter.com/1M2lImJdkB https://t.co/1M2lImJdkB October 12, 2026 https://x.com/finkd/status/2109660829952176518?ref src=twsrc%5Etfw vibes kinda good 🍉 — Alexandr Wang @alexandr wang October 12, 2026 https://x.com/alexandr wang/status/2109657401825861940?ref src=twsrc%5Etfw — Matt Deitke @mattdeitke October 12, 2026 https://x.com/mattdeitke/status/2109667603405050326?ref src=twsrc%5Etfw 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. Business Insider https://www.businessinsider.com/meta-launches-muse-wonder-series-to-rival-anthropic-openai-models-2026-10 Get AI news in your inbox Daily digest of what matters in AI. Key Terms Explained Anthropic https://www.machinebrief.com/glossary/anthropic An AI safety company founded in 2021 by former OpenAI researchers, including Dario and Daniela Amodei. Compute https://www.machinebrief.com/glossary/compute The processing power needed to train and run AI models. OpenAI https://www.machinebrief.com/glossary/openai The AI company behind ChatGPT, GPT-4, DALL-E, and Whisper. Prompting https://www.machinebrief.com/glossary/prompting The text input you give to an AI model to direct its behavior.