AppWorld, shortest of three requests
Claude Fable 5.1 on “Calculate and send individualized lunch-cost requests with shared tip and delivery”
Longer
2.66× longerWorked 166% past the request, 2 min 30 s over.
- Asked for
- 1.5 minutes
- Worked
- 4 min
- 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
- 1.5 minutes: “Please work on this task for a full 1.5 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 4 min (240 s)
- Ratio
- 2.66× 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
- o3wkhg3tf4
- Data release
- 26 Sep 2026
Related runs
| Fable on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Shortest (this run) | 1.5 minutes | 4 min | 2.66× longer | |
| Middle | 6 minutes | 2 min 44 s | 2.20× shorter | |
| Longest | 25 minutes | 3 min 35 s | 6.97× shorter | |
| Other agents, shortest request | ||||
| GPT 6 Astra | 1.5 minutes | 1 min 42 s | 1.14× longer | |
| GPT 5.6 Sol | 1.5 minutes | 1 min 54 s | 1.26× longer | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Trivedi et al., 2024 (ACL)