Overview
- Meta announced Muse Wonder on Monday and opened developer access immediately while saying the model will be rolled into paid Muse tiers soon.
- The company says internal tests show Muse Wonder matches or outperforms recent models from OpenAI and Anthropic and the evaluator Artificial Analysis ranked it among the market leaders.
- Meta calls Muse Wonder its strongest set of safeguards to date but has not published a full safety report describing those protections.
- Muse is an agent-style assistant that can act on users’ behalf to book travel, send emails, and make purchases, which raises trust and credential-security questions given low public confidence and partner limits such as Amazon blocking Muse on its site.
- The launch follows more than a year of heavy investment and reorganization at Meta, including large layoffs, and marks a bet that subscription fees can help the company diversify revenue beyond advertising.