METR Public Tasks

Evals

Benchmark / dataset

31 public autonomy tasks across 10 families plus summary of 100 private.

Category
Autonomy, R&D & self-replication
Insurance exposure
Agent actions
Publisher
METR
Publisher type
Nonprofit / research org
License
MIT
Status
Maintained (<12mo)
Last commit
2026-07-28
Stars
136
Notes
HCAST itself is private (180+ tasks).

View source on GitHub →

Wanich Risk Lab, 2026.