Meta releases Muse Ember, a new lightweight model for agent execution

Covered by 2 sourcesPublished Nov 9, 2026

Meta Superintelligence Lab launched Muse Ember, its lightest and most cost-efficient AI model, designed for complex reasoning and high-volume agent tasks across text, images, and audio. Ember offers lower cost per task than competitors and complements larger models like Muse Wonder for scalable AI agent workflows.

Also covered by 1 source

Vercel

What happened

Meta's Superintelligence Lab launched Muse Ember, a new lightweight AI model, on Monday. Muse Ember is now available through AI Gateway and is designed for tasks such as coding, document analysis, and agent execution. The model can process text, images, and audio, and is intended to handle large volumes of repetitive agent work. Muse Ember is positioned to work alongside larger models in scalable workflows.

Key details

Muse Ember costs $0.10 per million input tokens.
The model supports text, image, and audio inputs.
It is integrated with coding agents and APIs through AI Gateway.
Muse Ember is the fourth model in Meta's Muse lineup, after Spark, Glimmer, and Wonder.
It is designed to be more cost-efficient than competitors like Anthropic's Haiku and OpenAI's Luna.

Why it matters

Companies and developers using AI agents for high-volume tasks can now access a lower-cost, multi-modal model through AI Gateway. This reduces the operational cost for businesses handling large-scale coding, document processing, or support tasks. The availability of Muse Ember gives organizations a new option for integrating efficient, scalable AI into their workflows, potentially lowering expenses compared to other commercial and open-source models.