Introducing Our Smallest, Most Efficient Model: Muse Ember
Muse Ember sits on Artificial Analysis’s efficiency frontier: no model scores higher for a lower cost per task.
Meta’s model targets repeated agent work, including tool calls, coding subtasks, extraction and long-document parsing. It costs $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 200,000 tokens; above that, rates rise to $0.30 and $1.50. Willison cautions that token prices alone miss differences in how many reasoning tokens models consume. In his six SVG tests, higher reasoning settings did not consistently improve the drawings, and “high” failed to return a complete SVG.
Simon Willison's note
Meta’s model targets repeated agent work, including tool calls, coding subtasks, extraction and long-document parsing. It costs $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 200,000 tokens; above that, rates rise to $0.30 and $1.50. Willison cautions that token prices alone miss differences in how many reasoning tokens models consume. In his six SVG tests, higher reasoning settings did not consistently improve the drawings, and “high” failed to return a complete SVG.
Simon Willison's note
score 6