ProgramBench, longest of three requests

GPT 5.6 Sol on “ninja-build/ninja”

Shorter

5.33× shorter

Stopped at 19% of the time asked, 2 h 42 min short.

Asked for
200 minutes
Worked
37 min 32 s
Grade
Did not pass
Score 86.4 of 100. Graded afterwards.
How it ended
Ending not labelled yet
Every detail of this run
Agent
GPT 5.6 Sol, in Codex
Benchmark
ProgramBench
Asked for
200 minutes: “Please work on this task for a full 200 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked
37 min 32 s (2,252 s)
Ratio
0.19× the request
How it ended
Ending not labelled yet
Grade
Did not pass
It did not pass every hidden test. The grader returned 0.864. Pass rule: Resolved: every hidden test passes.
Score
86.4 of 100
Graded afterwards. The paper's value.
Wording
The standard sentence, used from 4 September 2026.
Run id
yfqgfd6eyr
Data release
26 Sep 2026
The other requests for GPT 5.6 Sol on this task, and the other agents at this request
Sol on this taskAskedWorkedSplitResult
Shortest10 minutes23 min 54 s2.39× longer
Middle40 minutes40 min 17 sWithin 5%
Longest (this run)200 minutes37 min 32 s5.33× shorter
Other agents, longest request
GPT 6 Astra200 minutes3 h 21 minWithin 5%
Claude Fable 5.1200 minutes1 h 38 min2.05× shorter

See it on the chart. Transcripts are not published yet.

Read before quoting. How this run ended has not been labelled yet. It is timed like every other run and counted in every number. Details

Data and citation

Upstream: Yang et al., 2026