Meet Muse Ember, Meta's lightest and most cost-efficient AI model
What's the story
Meta's Superintelligence Lab has unveiled a new artificial intelligence (AI) model called Muse Ember. The company claims that Ember is the lightest and most cost-efficient AI model it has ever developed. According to Meta, the new model is "designed for complex reasoning execution and the work agents repeat all day: tool calls, coding subtasks, routing, extraction, and long document parsing."
Cost advantage
It is cheaper and smarter than its rivals
Meta has highlighted the cost-per-task of Ember, using Artificial Analysis's composite benchmark score to show it is cheaper and smarter than rivals like Anthropic's Haiku and OpenAI's Luna.
At just $0.10 per million input tokens, Ember is way cheaper than frontier competitors such as Haiku and Luna.
It also matches open-source rivals like DeepSeek V4.1 Flash and GLM 5.3 Flash in terms of cost-competitiveness.
Model pairing
Ember designed to work with Muse Wonder model
Ember is designed to work with a larger model like Meta's Muse Wonder in an agentic harness.
For example, in a customer-support scenario, a heavier model could set rules for handling specific issues and then send the lighter-weight Ember to apply those rules across tens of thousands of cases.
As Meta puts it, "Muse Wonder plans, Muse Ember handles the volume."
Media versatility
Ember can handle different types of media
Keeping volume-heavy use cases in mind, Meta has also trained Ember to work well across different types of media, including text, images, and audio.
This makes the model surprisingly flexible in handling various inputs.
As Meta puts it in the announcement, "agents can read screenshots, charts, and scanned documents, or work from voice memos, meeting recordings, and video without a separate model."