Meta launches its first frontier AI model, Muse Wonder

Meta spent 14 months building a real rival to ChatGPT and Claude. On Monday, it showed the result: two new AI models, Muse Wonder and Muse Wonder Full. They come from Meta Superintelligence Labs, the team Mark Zuckerberg built after an aggressive hiring push. Meta says they're built to handle real personal and professional tasks, not just chat.

Meta is calling these 'frontier' models, the AI industry's term for systems at the very edge of what's possible. But that label is Meta's own. An independent scorer, Artificial Analysis, puts Muse Wonder Full at 56 points to Claude Opus 5.5's 58. That's essentially tied, with Anthropic still slightly ahead.

Keep reading — what to watch

Will Meta's own agents actually beat ChatGPT at real work, not just benchmark scores? Meta is betting the models' edge shows up on everyday personal and professional tasks, not lab tests. If regular people start trusting it with that kind of task, Meta finally has real footing in the AI race. It's chased that footing for two years.