Meta’s Muse Ember makes routine agent work cheaper to test
Meta’s new Muse Ember starts at $0.10 per million input tokens and $0.50 per million output tokens, offering a low-cost candidate for repeated tool calls, extraction and coding subtasks.
Released on November 9 through Meta’s model API, Muse Ember targets the small operations that accumulate inside agent workflows: routing, extracting information, calling tools and making bounded code changes.
For prompts up to 200,000 tokens, input and output cost $0.10 and $0.50 per million tokens; longer prompts cost $0.30 and $1.50. This is a proprietary model, rather than a release of downloadable weights. [1][2]
Artificial Analysis reports a rounded Intelligence Index score of 36 at Max, with an average cost of $0.14 per index task. That is a different capability and cost tier from Carbon High’s 60 and $7.76, or Fable 5.2 Max’s 65 and $26.54. The comparison supports testing a division of labor; it does not show that Ember can replace either frontier model on difficult work. [2][3][4]
Sources & details 4 sources
Discuss in your voice app
Take the reporting, visual explanations and original sources into a conversation. Copy the brief, paste it into your preferred assistant, then start its voice mode.
Paste the brief into your voice app to discuss this reporting.Reporting sources
- Meta — introducing Muse Ember
November 9 release, API prices, intended tasks and task-specific evaluation differences; vendor-reported comparisons remain qualified.
- Artificial Analysis — Muse Ember Max
External v5.0.1 index score, average task cost and proprietary-model status.
- Artificial Analysis — Carbon High
Same index and weighted task-cost framework, using regular API prices.
- Artificial Analysis — Fable 5.2 Max
Same index; Max configuration with default fallback. Costs do not imply equal success on every task.
Images & figures
A cheaper model can handle a different tier of work; costs use the same evaluation framework.
Euangelion · Source · Usage terms