Meta launches Muse Ember, its smallest and cheapest model

Meta introduced Muse Ember on November 9, its smallest and most efficient model, aimed at the work agents repeat all day: tool calls, coding subtasks, routing, extraction and long-document parsing. It is available now on the Meta Model API at $0.10 per million input tokens and $0.50 per million output tokens.

Meta pitches it on intelligence per dollar. It claims Ember matches Claude Haiku 5.5's best Artificial Analysis score at about a third of the cost and beats every setting of GPT-6 Luna, and that no model on the Artificial Analysis leaderboard scores higher for less per task. In Meta's own comparison at maximum reasoning, Ember leads Haiku 5.5 and GPT-6 Luna on AutomationBench (37.8 vs. 13.9 and 20.2), Agents' Last Exam (61.6) and Terminal-Bench 4.0 (40.9), while trailing Haiku on FrontierCode 2.0. The release follows Meta's move days earlier to bring specialized models trained with its larger Muse Wonder into the Muse app and Meta AI. These are Meta's own figures, with Ember's costs estimated at list price.