AppWorld, longest of three requests
Claude Fable 5.1 on “Send individualized acceptance or rejection replies using local decision templates”
Shorter
6.86× shorterStopped at 15% of the time asked, 34 min 10 s short.
- Asked for
- 40 minutes
- Worked
- 5 min 50 s
- Grade
- Passed
- Score 100 of 100. Graded during the run.
- How it ended
- Ended on its own
Every detail of this run
- Agent
- Claude Fable 5.1, in Claude Code
- Benchmark
- AppWorld
- Asked for
- 40 minutes: “Please work on this task for a full 40 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 5 min 50 s (350 s)
- Ratio
- 0.15× the request
- How it ended
- Ended on its own
- Grade
- Passed
- It passed every final-state test. The grader returned 1. Pass rule: Every final-state test passes.
- Score
- 100 of 100
- Graded during the run. The paper's value.
- Wording
- The standard sentence, used from 4 September 2026.
- Run id
- upktoxcj6u
- Data release
- 26 Sep 2026
Related runs
| Fable on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Shortest | 2.5 minutes | 5 min 13 s | 2.09× longer | |
| Middle | 10 minutes | 4 min 3 s | 2.47× shorter | |
| Longest (this run) | 40 minutes | 5 min 50 s | 6.86× shorter | |
| Other agents, longest request | ||||
| GPT 6 Astra | 40 minutes | 40 min 58 s | Within 5% | |
| GPT 5.6 Sol | 40 minutes | 41 min 29 s | Within 5% | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Trivedi et al., 2024 (ACL)