Meta has released Muse Ember, which it describes as "our smallest and most efficient model designed for complex reasoning execution," in a November 9 post on its research blog. The model is aimed at the repetitive work inside agent systems: tool calls, coding subtasks, routing, extraction and long-document parsing.
Muse Ember is available in Muse Code and through the Meta Model API under the model ID muse-ember-1.0. Meta says it is also offered on Google Cloud and Oracle Cloud and supports the Responses API. It has a 1 million token context window, can output up to 512,000 tokens, and accepts text, image, audio and video input.
For prompts up to 200,000 tokens, it costs $0.10 per million input tokens, $0.004 per million cached input tokens and $0.50 per million output tokens. Longer prompts cost $0.30, $0.012 and $1.50 respectively.
Meta's benchmark table compares Ember at maximum reasoning against Claude Haiku 5.5 and GPT-6 Luna. Meta reports 37.8% on AutomationBench-AA, against 13.9% for Haiku 5.5 and 20.2% for Luna, and 40.9% on Terminal-Bench 4.0, against 39.2% and 16.4%. On OSWorld 2.1 with partial credit, Ember scores 69.5% to Haiku's 68.9%. Ember trails Haiku 5.5 on FrontierCode 2.0-AA, at 27.2% against 34.9%, and ties Luna on GDP.pdf-AA at 22.8%. Meta says Ember matches Haiku's best Artificial Analysis Index score at about a third of the cost. These are Meta's own figures.
Meta frames Ember as the execution layer under larger Muse models, with a bigger model planning and Ember handling the volume. For developers running agent pipelines with many cheap calls, it is another low-cost option in the same price tier as other labs' smallest models. Meta's announcement does not say whether the model weights will be released.