Meta releases Muse Wonder models, claiming frontier status at lower cost
Meta has launched Muse Wonder and Muse Wonderful, its first frontier models, which score competitively on intelligence indices. The release follows a restructuring of Meta's AI efforts and positions the company as a fourth major frontier player.
Meta has released Muse Wonder and Muse Wonderful, its first frontier models, developed by its Super Intelligence Lab. Mark Zuckerberg claims the release places the company at the AI frontier alongside OpenAI and Anthropic. On the Artificial Analysis Intelligence Index, Muse Wonder scores 56 points, matching Anthropic's Sonnet and approaching Opus 5.5 at 58 points. The models are priced approximately 25% lower than Opus 5.5, with Muse Wonder costing $4 per million input tokens and $20 per million output tokens via the MetaModel API. Muse Wonderful is exclusive to Muse Code subscribers and is not available via API. Meta's stock rose 3.5% in after-hours trading following the announcement.
Independent tests show the models perform at the level of leading competitors in coding and agent tasks, though they lag significantly in knowledge questions and creative design. Reviewers noted that while Muse Wonder excels in agentic workflows such as billing ticket resolution and inbox triage, it underperforms in creative tasks, producing designs described as generic and behind competitors. The models feature "Active Memory" for long-term information retention, scoring higher on MemoryBench than Fable, Opus, or Astra. The release follows the August restructuring of Meta's AI efforts by ScaleAI founder Alexander Wang and was trained at Meta's Prometheus Data Center in central Ohio. Meta acknowledged that the model may reach a high risk threshold for chemical-biological misuse without safeguards.