Data source
METR Time Horizons
METR measures how long a task, timed in human working hours, an AI model can complete with a 50% success rate. We keep only these estimates, not its files, to calculate a score and quote the highest one.
Used to score: AI task length.
Provides measurements for the scored signals linked below.
Figure quoted from METR Time Horizons.
Visit the sourceBackground for: AI coding benchmarksLicense: No data license found
Highest estimate17 hours
What the data shows
Computed from every record, as of 2026-10.
- METR estimates the longest task a model completes half the time at about 17 hours of human work, with a range of 8 to 55 hours.
Limitations
- The newest estimates have wide ranges, and METR has said they depend heavily on modelling assumptions.
- METR did not measure every model, and measured some long after release.