Terminal-Bench 4.0, AgentTime v1.1, asked for four times its own forecast
GPT 6.1 Sol on “Complete a German month-end Intrastat filing across services”
Within 5%
Within 5%Worked 100% of the time asked, within 5% of it.
- Asked for
- 140 minutes
- Four times its own forecast of 35 min.
- Worked
- 2 h 21 min
- Grade
- Not graded
- How it ended
- Ended on its own
Every detail of this run
- Agent
- GPT 6.1 Sol, in Codex
- Benchmark
- Terminal-Bench 4.0
- Asked for
- 140 minutes, four times its own forecast of 35 min: “Please work on this task for a full 140 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 2 h 21 min (8,437 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
- x4o5mydehh
- Data release
- 2 Nov 2026
Related runs
| Sol 6.1 on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Quarter | 8.75 minutes | 9 min 15 s | 1.06× longer | |
| Four times (this run) | 140 minutes | 2 h 21 min | Within 5% | |
| Other agents, four times their own forecast | ||||
| Claude Opus 5.5 | 220 minutes | 46 s | 285.71× shorter | |
| Gemini 4 Argon | 72 minutes | 1 h 40 min | 1.39× longer | |
| MiMo V2.6 Pro | 440 minutes | 27 min 6 s | 16.24× shorter | |
| MiMo V2.6 Pro UltraSpeed | 320 minutes | 23 min 27 s | 13.65× shorter | |
| Muse Wonder 1.0 | 360 minutes | 27 min 20 s | 13.17× shorter | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Merrill et al., 2026 (ICLR)