METR Time Horizons

Evals

Benchmark / dataset

Data and code behind the 50% task-completion time-horizon metric.

Category
Autonomy, R&D & self-replication
Insurance exposure
Correlated/systemic
Publisher
METR
Publisher type
Nonprofit / research org
License
unknown
Status
Maintained (<12mo)
Last commit
2026-03-06
Stars
338
Notes
Widely cited capability trendline.

View source on GitHub →

Wanich Risk Lab, 2026.