Meta launches Muse Wonder models for AI agents
publishers 35articles 45first reported 6 Oct, 10:12 UTCdeveloping since 6 Oct · 21 editions
Meta launched Muse Wonder and Muse Wonder Full on Monday, introducing the first frontier models from its reorganized Superintelligence Labs division. The larger models follow the lightweight Muse Spark and are intended to power Meta’s Muse AI agent.
Both models are available to Muse Code subscribers. Wonder is also offered through the Meta Model API at $4 per million input tokens and $20 per million output tokens, while Wonder Full is not available through the API. Meta said Wonder will begin powering Muse for subscribers, but did not provide a timeline for a broader rollout.
Meta says the models perform well on complex tasks including coding and can compete with leading systems from OpenAI and Anthropic. Those comparisons are based on Meta’s internal benchmark testing and have not yet been confirmed through broad independent evaluation.
The company emphasized two capabilities designed for long-running personal-assistant tasks. “Active Memory” is intended to retain and synthesize user information over extended periods, while “Active Focus” is designed to help agents persist until tasks are completed. Meta reported that Wonder outperformed several rival models on the MemoryBench test.
The models were trained at Meta’s Prometheus data center in central Ohio, according to TechCrunch. Meta said they represent the next stage of its effort to build personalized AI agents, following the recent launches of Muse Spark and the Muse app.
HOW THIS STORY WAS MADE
Written from 8 articles, 4 with the publisher's own text; 8 independent newsrooms once syndicated copies count as one; this version written 168 h after the record first saw the story; the editor kept it as written.
- Publishers
- 35
- Source articles
- 45
- Given to the writer
- 8, 4 with the article's own text
- Independent newsrooms
- 8
- This version written
- 168 h after the record first saw the story
- Second model (editor)
- kept as written
- Publication gate
- passed
- Human review
- none
WHAT THE EDITOR DID
The second model checked the writer's draft against the same sources and published it as written.
Written by a language model from the sources above, then checked by a second model that may only cut, attribute or correct. How it works →