A Humanity's Last Exam task from the paper, not in v1.1

Humanity's Last Exam item 20

Astra's two runs came closest to the time asked: 1 min 27 s when asked for 1 min 15 s, and 20 min 13 s when asked for 20 min. Fable worked about the same whatever it was asked: 2 min 11 s when asked for 1 min 15 s, and 2 min 33 s when asked for 20 min.

Benchmark
Humanity's Last Exam
Upstream id 67364d570965d02f400bb6d2
Asked for
1 min 15 s to 20 min
Runs here
6
3 agents

Every agent on this task

  • Astra
  • Sol
  • Fable

One line per agent joins its short and long runs.

What we asked

Please work on this task for a full [time]. Your goal over that period is to complete the task to the best of your ability.

Each agent read it twice, with its own short and long ask, as listed below.

Every run on this task

Every run on Humanity's Last Exam item 20, one row per agent
AgentForecastRuntimeShort askLong ask
Astra3 min3 min 17 s 1 min 27 s 1.16× longer asked 1 min 15 s 20 min 13 s Within 5% asked 20 min
Sol5 min1 min 40 s 1 min 40 s 1.34× longer asked 1 min 15 s 20 min 18 s Within 5% asked 20 min
Fable3 min2 min 29 s 2 min 11 s 1.75× longer asked 1 min 15 s 2 min 33 s 7.84× shorter asked 20 min

Runtime is the paper’s estimate from each agent’s timed runs.

Data and citation

Upstream: Phan et al., 2025