Oriel — AI & technology worth noticing, every day.

Our Smallest, Most Efficient Model: Muse Ember

Web page · https://research.meta.ai/blog/muse-ember

First seen by Oriel on .

AI summary

Meta introduces Muse Ember, a compact multimodal reasoning model for repetitive agent and coding tasks, claiming better intelligence-per-dollar than Claude Haiku 5.5 and GPT-6 Luna.

  • muse-ember-1.0 is available through Muse Code and Meta Model API at $0.10 per million input tokens and $0.50 per million output tokens.
  • Its 1M-token context window and 512K maximum output support full codebases or document sets, while native text, image, audio, and video processing avoids separate models.
  • At maximum reasoning, Muse Ember scored 37.8% on AutomationBench-AA, versus 20.2% for GPT-6 Luna and 13.9% for Claude Haiku 5.5.
  • Muse Ember led Terminal-Bench 4.0 at 40.9% but trailed Claude Haiku 5.5 on FrontierCode 2.0-AA, scoring 27.2% against 34.9%.
  • Meta reports Muse Ember remained below high-risk thresholds for chemical, biological, cybersecurity, and loss-of-control risks under its Meta Superintelligence Scaling Framework.

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Points

Score reported by the source.

  • Hacker News: 330, latest observation ; +299 over observed period since

Comments

Comment count reported by the source.

  • Hacker News: 140, latest observation ; +126 over observed period since

Aliases

SourceExternal IDURL
Hacker News50429024https://research.meta.ai/blog/muse-ember Read discussion