AssistantBench, AgentTime v1.1, asked for four times its own forecast
Claude Opus 5.5 on “AssistantBench validation task 18”
Shorter
2.50× shorterStopped at 40% of the time asked, 35 min 59 s short.
- Asked for
- 60 minutes
- Four times its own forecast of 15 min.
- Worked
- 24 min 2 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
- AssistantBench
- Asked for
- 60 minutes, four times its own forecast of 15 min: “Please work on this task for a full 60 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 24 min 2 s (1,442 s)
- Ratio
- 0.40× 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
- uim4vp4vmy
- Data release
- 2 Nov 2026
Related runs
| Opus 5.5 on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Quarter | 3.75 minutes | 13 min 31 s | 3.60× longer | |
| Four times (this run) | 60 minutes | 24 min 2 s | 2.50× shorter | |
| Other agents, four times their own forecast | ||||
| Gemini 4 Argon | 40 minutes | 40 min 39 s | Within 5% | |
| GPT 6.1 Sol | 88 minutes | 1 h 29 min | Within 5% | |
| MiMo V2.6 Pro | 100 minutes | 55 min 28 s | 1.80× shorter | |
| MiMo V2.6 Pro UltraSpeed | 120 minutes | 6 h | 3.00× longer | |
| Muse Wonder 1.0 | 88 minutes | 1 h 29 min | Within 5% | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yoran et al., 2024 (EMNLP)