ProgramBench, middle of three requests
GPT 5.6 Sol on “cweill/gotests”
Within 5%
Within 5%Worked 102% of the time asked, within 5% of it.
- Asked for
- 50 minutes
- Worked
- 50 min 59 s
- Grade
- Did not pass
- Score 87.7 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
- cweill/gotests
- Asked for
- 50 minutes: “Please work on this task for a full 50 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 50 min 59 s (3,059 s)
- Ratio
- 1.02× the request
- How it ended
- Ended on its own
- Grade
- Did not pass
- It did not pass every hidden test. The grader returned 0.877. Pass rule: Resolved: every hidden test passes.
- Score
- 87.7 of 100
- Graded afterwards. The paper's value.
- Wording
- The standard sentence, used from 4 September 2026.
- Run id
- 646yz3joqq
- Data release
- 26 Sep 2026
Related runs
| Sol on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Shortest | 12.5 minutes | 19 min 48 s | 1.58× longer | |
| Middle (this run) | 50 minutes | 50 min 59 s | Within 5% | |
| Longest | 200 minutes | 1 h 11 min | 2.81× shorter | |
| Other agents, middle request | ||||
| GPT 6 Astra | 50 minutes | 50 min 38 s | Within 5% | |
| Claude Fable 5.1 | 50 minutes | 1 h 13 min | 1.46× longer | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yang et al., 2026