Meta Launches Muse Wonder Models, Claiming State-of-the-Art Performance in AI Agents and Coding

Meta has unveiled Muse Wonder and Muse Wonder Full, frontier-class AI models designed to power agentic tasks and long-horizon coding work, with performance claims that position the company competitively against OpenAI and Anthropic. Announced October 12, Meta describes the models as its most capable to date, delivering what it calls state-of-the-art results on benchmarks measuring personal agents, memory-intensive workflows, and software engineering tasks.