Meta launched Muse Ember, its smallest and most efficient model for routine agent work. The company priced the model at $0.10 per million input tokens and $0.50 per million output tokens. Muse Ember is available today on the Meta Model API.
Cost and efficiency
Meta said the model matches Claude Haiku 5.5's best score at about a third of the cost. It also beat every GPT-6 Luna setting on efficiency. Artificial Analysis found no model scores higher for less per task.
Agent workflows
Muse Ember handles tool calls, coding subtasks, routing, extraction, and long document parsing. It can carry a business workflow from end to end. It pulls records, calls tools in sequence, and updates systems across apps.
Model specs
The model has a 1M-token context window and up to 512K output tokens. It can take in a full codebase or document set in one call. It also processes image, audio, and video content.
Developers can also access Muse Ember through Muse Code, Google Cloud, and Oracle Cloud. Early partners tested the model on contract reviews and coding tasks. It is built for small, structured steps agents repeat all day.