← Blog

AI tools2026-11-10Source research.meta.ai

What is Muse Ember? Meta's smallest, most efficient agent model

Muse Ember, launched by Meta on November 9, 2026, is a lightweight model for the work agents repeat all day: tool calls, coding subtasks, routing, extraction and long document parsing.

Meta introduced Muse Ember on November 9, 2026 as the smallest and most efficient model in the Muse family. It is built for the steps agents repeat all day, such as tool calls, coding subtasks, routing requests, extracting fields and parsing long documents. Meta says it is the first specialized model trained using Muse Wonder Full.

Muse Ember has a 1M token context window and up to about 512K output tokens. It natively reads text, images, audio, video and PDF, so it can work from screenshots, charts and scanned documents and can operate a computer from screenshots. On the Meta Model API it is muse-ember-1.0, priced at 0.10 dollars per million input tokens and 0.50 dollars per million output tokens up to 200K input tokens, with higher long context rates above that.

Meta says Muse Ember sits on the efficiency frontier of the Artificial Analysis leaderboard and matches the best Claude Haiku 5.5 score at about a third of the cost. The suggested pattern is to let Muse Spark or Muse Wonder plan the work while Muse Ember handles the volume.

Muse Ember is available on every Muse Code plan, on the Meta Model API, and on Google Cloud and Oracle Cloud. For website builders, repetitive jobs such as writing image alt text, cleaning product data or rewriting copy across many pages are a natural fit for a low cost model like this.