Artificial intelligence / news brief
Meta launches Muse Ember for high-volume agent tasks
- Event
- Published

Meta released Muse Ember on Muse Code and its Model API at $0.10 per million input tokens and $0.50 per million output tokens. The multimodal model has a 1M-token context window and is aimed at tool calls, routing, extraction and coding subtasks. Meta's efficiency and benchmark comparisons are vendor-reported; Ember trails Claude Haiku 5.5 on Meta's cited FrontierCode result.
Why it matters
A low-cost multimodal model designed for repetitive agent steps could reduce workflow costs, but independent quality and reliability testing is still missing.
Sources
- Introducing Our Smallest, Most Efficient Model: Muse Ember (external site)Meta AI Research · primary source ·
- Meta releases Muse Ember, a new lightweight model for agent execution (external site)TechCrunch · coverage ·
- Meta is launching its smallest and most efficient AI model: Muse Ember (external site)The Verge · coverage ·