Terminal-Bench 4.0, AgentTime v1.1, asked for four times its own forecast
Claude Opus 5.5 on “hof-topology-interpenetration”
Shorter
8.19× shorterStopped at 12% of the time asked, 2 h 56 min short.
- Asked for
- 200 minutes
- Four times its own forecast of 50 min.
- Worked
- 24 min 25 s
- Grade
- Not graded
- How it ended
- Ended on its own
Every detail of this run
- Agent
- Claude Opus 5.5, in Claude Code
- Benchmark
- Terminal-Bench 4.0
- Asked for
- 200 minutes, four times its own forecast of 50 min: “Please work on this task for a full 200 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 24 min 25 s (1,465 s)
- Ratio
- 0.12× 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
- vr2tiw7uhv
- Data release
- 2 Nov 2026
Related runs
| Opus 5.5 on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Quarter | 12.5 minutes | 7 min 38 s | 1.64× shorter | |
| Four times (this run) | 200 minutes | 24 min 25 s | 8.19× shorter | |
| Other agents, four times their own forecast | ||||
| Gemini 4 Argon | 112 minutes | 26 min 7 s | 4.29× shorter | |
| GPT 6.1 Sol | 240 minutes | 4 h 1 min | Within 5% | |
| MiMo V2.6 Pro | 18 hours | 1 h 17 min | 14.00× shorter | |
| MiMo V2.6 Pro UltraSpeed | 28 hours | 28 min 29 s | 58.99× shorter | |
| Muse Wonder 1.0 | 760 minutes | 24 min 28 s | 31.07× shorter | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Merrill et al., 2026 (ICLR)