Meta releases Muse Ember, a lightweight model for agent work
Original title: Meta releases Muse Ember, a new lightweight model for agent execution
Meta prices Ember at 10 cents per million input tokens and pairs it with Muse Wonder for volume.
Meta Superintelligence Lab released Muse Ember, calling it its lightest and most cost-efficient model, for tool calls, coding subtasks, routing, extraction, and long-document parsing. Meta says Ember is cheaper than Anthropic’s Haiku and OpenAI’s Luna and cost-competitive with DeepSeek V4.1 Flash and GLM 5.3 Flash, citing Artificial Analysis’s composite score. It is meant to pair with Muse Wonder, which plans while Ember handles volume, and accepts text, images, and audio. Ember is the fourth Muse-branded model.
Unconfirmed: Cost and quality comparisons are Meta’s claims using Artificial Analysis’s score, not independently confirmed here.
- Input price is 10 cents per million tokens.
- Meta’s comparison claims Ember is cheaper per task than Haiku and Luna.
- It accepts text, images, and audio, including screenshots and voice memos.
- Earlier Muse models are Spark, open-source Glimmer, and Wonder.
- Vendors
- Meta
- Organizations
- Meta Superintelligence Lab
- AI models
- Muse EmberMuse Wonder
From the original article
2 min read at techcrunch.com
At 10 cents per million input tokens, Ember is significantly cheaper than frontier competitors like Haiku and Luna.
By Russell Brandom · techcrunch.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.