Meta introduces Muse Ember to automate repetitive tasks

Meta launches Muse Ember, a compact model designed for the repetitive tasks of agents. It is offered via API with a very low announced pricing and a context window of one million tokens. Initial cost benchmarks and scores position it competitively, with limitations on real coding tasks compared to Claude Haiku.

Immediate Access and Published Pricing for the API

Muse Ember is accessible via the Meta Model API and Muse Code. The pricing is set at $0.10 per million input tokens and $0.50 per million output tokens. The model can handle up to one million tokens per request and supports text, images, audio, and video.

Cost per Task and Measured Gaps Against Competitors

In the Artificial Analysis test, Muse Ember costs about $0.14 per task, while Muse Spark 1.3 reaches $2.39 for a score of 36 against 37. On the FrontierCode 2.0-AA benchmark, Claude Haiku 5.5 shows a success rate of 34.9% on real coding tasks, while Muse Ember achieves 27.2%. Meta indicates that Muse Ember delivers more capabilities per dollar at every reasoning level than Claude Haiku 5.5. Muse Ember is also compared to GPT-6 Luna, and it ranks first in the majority of its internal evaluations.

Targeting Agent Tasks and Model Positioning

Meta has launched Muse Ember, which it describes as its most compact and high-performing model to date. This model targets agent use cases such as tool calling, performing small programming tasks, query guidance, information extraction, and processing large documents.