Humanity's Last Exam, AgentTime v1.1, asked for four times its own forecast
Muse Wonder 1.0 on “Humanity's Last Exam item 28”
Shorter
2.70× shorterStopped at 37% of the time asked, 15 min 6 s short.
- Asked for
- 24 minutes
- Four times its own forecast of 6 min.
- Worked
- 8 min 54 s
- Grade
- Not graded
- How it ended
- Ended on its own
Every detail of this run
- Agent
- Muse Wonder 1.0, in Muse Code
- Benchmark
- Humanity's Last Exam
- Asked for
- 24 minutes, four times its own forecast of 6 min: “Please work on this task for a full 24 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 8 min 54 s (534 s)
- Ratio
- 0.37× 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
- yd4eidqkhd
- Data release
- 2 Nov 2026
Related runs
| Muse Wonder on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Quarter | 1.5 minutes | 3 min 53 s | 2.59× longer | |
| Four times (this run) | 24 minutes | 8 min 54 s | 2.70× shorter | |
| Other agents, four times their own forecast | ||||
| Claude Opus 5.5 | 24 minutes | 33 min 32 s | 1.40× longer | |
| Gemini 4 Argon | 600 seconds | 10 min 49 s | 1.08× longer | |
| GPT 6.1 Sol | 36 minutes | 36 min 35 s | Within 5% | |
| MiMo V2.6 Pro | 80 minutes | 37 min 29 s | 2.13× shorter | |
| MiMo V2.6 Pro UltraSpeed | 60 minutes | 5 min 4 s | 11.86× shorter | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Phan et al., 2025