ProgramBench, longest of three requests
GPT 5.6 Sol on “nuta/nsh”
Shorter
6.63× shorterStopped at 15% of the time asked, 2 h 50 min short.
- Asked for
- 200 minutes
- Worked
- 30 min 10 s
- Grade
- Did not pass
- Score 76.8 of 100. Graded afterwards.
- How it ended
- Ended on its own
Every detail of this run
- Agent
- GPT 5.6 Sol, in Codex
- Benchmark
- ProgramBench
- Task
- nuta/nsh
- 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
- 30 min 10 s (1,810 s)
- Ratio
- 0.15× the request
- How it ended
- Ended on its own
- Grade
- Did not pass
- It did not pass every hidden test. The grader returned 0.768. Pass rule: Resolved: every hidden test passes.
- Score
- 76.8 of 100
- Graded afterwards. The paper's value.
- Wording
- The standard sentence, used from 4 September 2026.
- Run id
- ro2xeyplfd
- Data release
- 26 Sep 2026
Related runs
| Sol on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Shortest | 12.5 minutes | 22 min 8 s | 1.77× longer | |
| Middle | 50 minutes | 49 min 19 s | Within 5% | |
| Longest (this run) | 200 minutes | 30 min 10 s | 6.63× shorter | |
| Other agents, longest request | ||||
| GPT 6 Astra | 200 minutes | 3 h 20 min | Within 5% | |
| Claude Fable 5.1 | 200 minutes | 6 h 40 min | 2.00× longer | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yang et al., 2026