ProgramBench, longest of three requests
Claude Fable 5.1 on “noborus/trdsql”
Shorter
2.00× shorterStopped at 50% of the time asked, 1 h 40 min short.
- Asked for
- 200 minutes
- Worked
- 1 h 40 min
- Grade
- Did not pass
- Score 93.9 of 100. Graded afterwards.
- How it ended
- Ended on its own
Every detail of this run
- Agent
- Claude Fable 5.1, in Claude Code
- Benchmark
- ProgramBench
- Task
- noborus/trdsql
- 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
- 1 h 40 min (6,008 s)
- Ratio
- 0.50× the request
- How it ended
- Ended on its own
- Grade
- Did not pass
- It did not pass every hidden test. The grader returned 0.939. Pass rule: Resolved: every hidden test passes.
- Score
- 93.9 of 100
- Graded afterwards. The paper's value.
- Wording
- The standard sentence, used from 4 September 2026.
- Run id
- l3hh2gbcb6
- Data release
- 26 Sep 2026
Related runs
| Fable on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Shortest | 12.5 minutes | 7 min 36 s | 1.64× shorter | |
| Middle | 50 minutes | 20 s | 147.06× shorter | |
| Longest (this run) | 200 minutes | 1 h 40 min | 2.00× shorter | |
| Other agents, longest request | ||||
| GPT 6 Astra | 200 minutes | 3 h 21 min | Within 5% | |
| GPT 5.6 Sol | 200 minutes | 1 h | 3.33× shorter | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yang et al., 2026