ProgramBench, AgentTime v1.1, asked for four times its own forecast
GPT 6.1 Sol on “typst/typst”
Within 5%
Within 5%Worked 100% of the time asked, within 5% of it.
- Asked for
- 420 minutes
- Four times its own forecast of 1 h 45 min.
- Worked
- 7 h
- Grade
- Not graded
- How it ended
- Ended on its own
Every detail of this run
- Agent
- GPT 6.1 Sol, in Codex
- Benchmark
- ProgramBench
- Task
- typst/typst
- Asked for
- 420 minutes, four times its own forecast of 1 h 45 min: “Please work on this task for a full 420 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 7 h (25,222 s)
- Ratio
- 1.00× the request
- How it ended
- Ended on its own
- Grade
- Not graded
- Clock tool
- The agent could call a wall-clock tool to read the current time.
- Wording
- The paper's standard sentence, sent in AgentTime v1.1.
- Run id
- 6vhy22d5lb
- Data release
- 2 Nov 2026
Related runs
| Sol 6.1 on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Quarter | 26.25 minutes | 26 min 37 s | Within 5% | |
| Four times (this run) | 420 minutes | 7 h | Within 5% | |
| Other agents, four times their own forecast | ||||
| Claude Opus 5.5 | 16 hours | 15 h 41 min | Within 5% | |
| Gemini 4 Argon | 440 minutes | 1 h 23 min | 5.29× shorter | |
| MiMo V2.6 Pro | 84 hours | 36 h 22 min | 2.31× shorter | |
| MiMo V2.6 Pro UltraSpeed | 80 hours | 1 h 22 min | 58.54× shorter | |
| Muse Wonder 1.0 | 120 hours | 22 min 37 s | 318.30× shorter | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yang et al., 2026