AssistantBench, middle of three requests
GPT 6 Astra on “AssistantBench validation task 5”
Longer
1.25× longerWorked 25% past the request, 1 min 14 s over.
- Asked for
- 5 minutes
- Worked
- 6 min 14 s
- Grade
- Did not pass
- Score 0 of 100. Graded afterwards.
- How it ended
- Ended on its own
Every detail of this run
- Agent
- GPT 6 Astra, in Codex
- Benchmark
- AssistantBench
- Asked for
- 5 minutes: “Please work on this task for a full 5 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 6 min 14 s (374 s)
- Ratio
- 1.25× 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. Pass rule: Exact match with the reference answer.
- Score
- 0 of 100
- Graded afterwards. The paper's value.
- Wording
- The standard sentence, used from 4 September 2026.
- Run id
- ukbtlljr6u
- Data release
- 26 Sep 2026
Related runs
| Astra on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Shortest | 1.25 minutes | 1 min 56 s | 1.55× longer | |
| Middle (this run) | 5 minutes | 6 min 14 s | 1.25× longer | |
| Longest | 20 minutes | 9 min 37 s | 2.08× shorter | |
| Other agents, middle request | ||||
| GPT 5.6 Sol | 5 minutes | 2 min 11 s | 2.29× shorter | |
| Claude Fable 5.1 | 5 minutes | 6 min 34 s | 1.31× longer | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yoran et al., 2024 (EMNLP)