METR task time horizon
Length of tasks models complete at 50% reliability; doubling about every 7 months
agents and tool useactive
Definition
- Version
- not in the notes
- Category
- agents and tool use
- Maintainer
- METR
- Status
- active
- Home page
- https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/
Sources
- Measuring AI Ability to Complete Long Software Tasks metr.org https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/
Scores
The notes record no score on this benchmark.
Where this comes from
Scores from different settings, harnesses or dates are not directly comparable; the row says which. Sorted by score, highest first, as a reading aid, not a ranking.
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "METR Task Time Horizon", https://aurion.semai.id/benchmarks/metr-time-horizon/, records of October 2026.