ProgramBench, longest of three requests
Claude Fable 5.1 on “facebook/zstd”
Shorter
2.47× shorterStopped at 40% of the time asked, 1 h 29 min short.
- Asked for
- 150 minutes
- Worked
- 1 h 1 min
- Grade
- Did not pass
- Score 73.1 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
- facebook/zstd
- Asked for
- 150 minutes: “Please work on this task for a full 150 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 1 h 1 min (3,644 s)
- Ratio
- 0.40× the request
- How it ended
- Ended on its own
- Grade
- Did not pass
- It did not pass every hidden test. The grader returned 0.731. Pass rule: Resolved: every hidden test passes.
- Score
- 73.1 of 100
- Graded afterwards. The paper's value.
- Wording
- The standard sentence, used from 4 September 2026.
- Run id
- vf6auyhpzl
- Data release
- 26 Sep 2026
Related runs
| Fable on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Shortest | 10 minutes | 8 min 50 s | 1.13× shorter | |
| Middle | 40 minutes | 20 min 18 s | 1.97× shorter | |
| Longest (this run) | 150 minutes | 1 h 1 min | 2.47× shorter | |
| Other agents, longest request | ||||
| GPT 6 Astra | 150 minutes | 2 h 32 min | Within 5% | |
| GPT 5.6 Sol | 150 minutes | 2 h 31 min | Within 5% | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yang et al., 2026