ProgramBench, longest of three requests
Claude Fable 5.1 on “lfos/calcurse”
Shorter
1.24× shorterStopped at 81% of the time asked, 1 h 17 min short.
- Asked for
- 400 minutes
- Worked
- 5 h 23 min
- Grade
- Did not pass
- Score 95.6 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
- lfos/calcurse
- Asked for
- 400 minutes: “Please work on this task for a full 400 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 5 h 23 min (19,381 s)
- Ratio
- 0.81× the request
- How it ended
- Ended on its own
- Grade
- Did not pass
- It did not pass every hidden test. The grader returned 0.956. Pass rule: Resolved: every hidden test passes.
- Score
- 95.6 of 100
- Graded afterwards. The paper's value.
- Wording
- The standard sentence, used from 4 September 2026.
- Run id
- ndwpwyd3xx
- Data release
- 26 Sep 2026
Related runs
| Fable on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Shortest | 25 minutes | 8 min 41 s | 2.88× shorter | |
| Middle | 100 minutes | 2 h 16 min | 1.36× longer | |
| Longest (this run) | 400 minutes | 5 h 23 min | 1.24× shorter | |
| Other agents, longest request | ||||
| GPT 6 Astra | 400 minutes | 6 h 41 min | Within 5% | |
| GPT 5.6 Sol | 400 minutes | 13 min 54 s | 28.79× shorter | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yang et al., 2026