Meta releases a small, cheap AI model for repetitive work

Meta has released Muse Ember, a small AI model built to be cheap. It is meant for the repetitive jobs that AI agents (programs that carry out tasks on their own) do all day. Meta says it gets you more for your money than Claude Haiku 5.5 or GPT-6 Luna.

What it does

Muse Ember is Meta's smallest and most efficient model so far. It is aimed at calling tools, handling small coding tasks, routing requests, pulling data out of text and processing long documents. It works with text, images, audio and video. It can take in up to one million tokens at once. A token is a small chunk of text.

What it costs

It is available now through the Meta Model API and Muse Code. Pricing is $0.10 per million input tokens and $0.50 per million output tokens.

In one Artificial Analysis test, Muse Ember cost about $0.14 per task. Meta's larger Muse Spark 1.3 cost $2.39 for nearly the same score, 36 versus 37.

The catch

Meta says Muse Ember leads in most of its own tests. Those are Meta's tests, so treat them with care. On real coding tasks in FrontierCode 2.0-AA, Claude Haiku 5.5 scored 34.9 percent and Muse Ember scored 27.2 percent.

The material does not say how the model performs on everyday business work. If you pay for AI tools to handle routine tasks, a cheaper model could lower your bill. Test it on your own tasks first.