AssistantBench, longest of three requests
GPT 5.6 Sol on “AssistantBench validation task 18”
Shorter
1.41× shorterStopped at 71% of the time asked, 5 min 48 s short.
- Asked for
- 20 minutes
- Worked
- 14 min 12 s
- Grade
- Did not pass
- Score 14 of 100. Graded afterwards.
- How it ended
- Ended on its own
Every detail of this run
- Agent
- GPT 5.6 Sol, in Codex
- Benchmark
- AssistantBench
- Asked for
- 20 minutes: “Please work on this task for a full 20 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 14 min 12 s (852 s)
- Ratio
- 0.71× the request
- How it ended
- Ended on its own
- Grade
- Did not pass
- It did not match the reference answer exactly. The grader returned 0.14. Pass rule: Exact match with the reference answer.
- Score
- 14 of 100
- Graded afterwards. The paper's value.
- Wording
- The standard sentence, used from 4 September 2026.
- Run id
- cy674fxwal
- Data release
- 26 Sep 2026
Related runs
| Sol on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Shortest | 1.25 minutes | 9 min 42 s | 7.77× longer | |
| Middle | 5 minutes | 9 min 43 s | 1.94× longer | |
| Longest (this run) | 20 minutes | 14 min 12 s | 1.41× shorter | |
| Other agents, longest request | ||||
| GPT 6 Astra | 20 minutes | 13 min 47 s | 1.45× shorter | |
| Claude Fable 5.1 | 20 minutes | 40 min | 2.00× longer | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yoran et al., 2024 (EMNLP)