BrowseComp, AgentTime v1.1, asked for four times its own forecast
MiMo V2.6 Pro UltraSpeed on “BrowseComp question 6”
Shorter
43.78× shorterStopped at 2% of the time asked, 2 h 9 min short.
- Asked for
- 132 minutes
- Four times its own forecast of 33 min.
- Worked
- 3 min 1 s
- Grade
- Not graded
- How it ended
- Ended on its own
Every detail of this run
- Agent
- MiMo V2.6 Pro UltraSpeed, in MiMo Code
- Benchmark
- BrowseComp
- Asked for
- 132 minutes, four times its own forecast of 33 min: “Please work on this task for a full 132 minutes. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 3 min 1 s (181 s)
- Ratio
- 0.02× 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
- o7fuug3xke
- Data release
- 2 Nov 2026
Related runs
| MiMo UltraSpeed on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Quarter | 8.25 minutes | 2 min 14 s | 3.70× shorter | |
| Four times (this run) | 132 minutes | 3 min 1 s | 43.78× shorter | |
| Other agents, four times their own forecast | ||||
| Claude Opus 5.5 | 720 seconds | 12 min 27 s | Within 5% | |
| Gemini 4 Argon | 20 minutes | 21 min 47 s | 1.09× longer | |
| GPT 6.1 Sol | 32 minutes | 32 min 42 s | Within 5% | |
| MiMo V2.6 Pro | 128 minutes | 28 min 10 s | 4.54× shorter | |
| Muse Wonder 1.0 | 72 minutes | 3 min 49 s | 18.90× shorter | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Wei et al., 2025