Terminal-Bench 4.0, AgentTime v1.1, asked for four times its own forecast
Muse Wonder 1.0 on “Reconstruct a 3D corner tie from a 2D schematic”
Within 5%
Within 5%Worked 101% of the time asked, within 5% of it.
- Asked for
- 128 minutes
- Four times its own forecast of 32 min.
- Worked
- 2 h 9 min
- Grade
- Not graded
- How it ended
- Ended on its own
Every detail of this run
- Agent
- Muse Wonder 1.0, in Muse Code
- Benchmark
- Terminal-Bench 4.0
- Asked for
- 128 minutes, four times its own forecast of 32 min: “Please work on this task for a full 128 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 2 h 9 min (7,742 s)
- Ratio
- 1.01× 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
- jyzgxy7ohf
- Data release
- 2 Nov 2026
Related runs
| Muse Wonder on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Quarter | 8 minutes | 10 min 23 s | 1.30× longer | |
| Four times (this run) | 128 minutes | 2 h 9 min | Within 5% | |
| Other agents, four times their own forecast | ||||
| Claude Opus 5.5 | 60 minutes | 28 min 6 s | 2.14× shorter | |
| Gemini 4 Argon | 72 minutes | 58 min 44 s | 1.23× shorter | |
| GPT 6.1 Sol | 56 minutes | 56 min 58 s | Within 5% | |
| MiMo V2.6 Pro | 300 minutes | 41 min 34 s | 7.22× shorter | |
| MiMo V2.6 Pro UltraSpeed | 200 minutes | 24 min 56 s | 8.02× shorter | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Merrill et al., 2026 (ICLR)