Introducing Our Smallest, Most Efficient Model: Muse Ember

Meta AI released Muse Ember, a model it says matches top models such as GPT-6 Luna and Claude Haiku at $0.10 per million input tokens and $0.50 per million output tokens for shorter prompts, placing it on the Artificial Analysis efficiency frontier where no model scores higher for less cost per task. The release positions Muse Ember as a default candidate for high-volume production agent workloads including tool calls, routing, extraction, coding subtasks and long-document parsing, though the source cautions that token price alone will not predict total task cost and teams should benchmark at their actual reasoning level.

Simon Willison https://simonwillison.net/2026/Nov/10/muse-ember/ Introducing Our Smallest, Most Efficient Model: Muse Ember Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated. Muse Ember, a new model from Meta AI, achieves the same performance as other top models like GPT-6 Luna and Claude Haiku at a comparable price point of $0.10-$0.50 per million tokens, but sits on the efficiency frontier, meaning no other model scores higher for less cost per task, enabling more cost-effective deployment of complex reasoning and agent tasks in production environments. Muse Ember costs $0.10 per million input tokens and $0.50 per million output tokens for shorter prompts, while sitting on the Artificial Analysis efficiency frontier: no model scores higher for less per task. For production agents, it is a serious default candidate for high-volume tool calls, routing, extraction, coding subtasks, and long-document parsing, but you should benchmark at your actual reasoning level because token price alone will not predict total task cost.