Updated daily at 05:00 UTC
Benchmark
METR Time Horizons
Measures: An agent attempts a suite of software-engineering, ML-research-engineering and cybersecurity tasks (HCAST, RE-Bench and short software tasks) whose length is calibrated by how long human experts take; a logistic fit gives the human task length at which the agent succeeds 50% of the time (the 50% time horizon), with an 80% horizon reported alongside.
- Maintainer
- METR
- Introduced
- 2025
- Status
- Active
- Saturation
- Near-ceiling
- Typical provenance
- Independent
- In scoring set
- No
Not scored: Open-ended minutes scale with no fixed ceiling; METR says values above 16 hours are unreliable with its current tasks and has added no model since 2026-05-08.
Leaderboard
Every result we hold, whoever ran it. Only independent and third-party rows count towards a score.
No results to show.