Task 6 of 12 in BrowseComp
BrowseComp question 6
Sol 6.1's two runs came closest to the time asked: 2 min 19 s when asked for 2 min, and 32 min 42 s when asked for 32 min. MiMo UltraSpeed worked about the same whatever it was asked: 2 min 14 s when asked for 8 min 15 s, and 3 min 1 s when asked for 2 h 12 min.
- Asked for
- 45 s to 2 h 12 min
- Runs here
- 12
- 6 agents
Every agent on this task
- Opus 5.5
- Argon
- Sol 6.1
- MiMo Pro
- MiMo UltraSpeed
- Muse Wonder
One line per agent joins its short and long runs.
What we asked
Please work on this task for a full [time]. Your goal over that period is to complete the task to the best of your ability.
Each agent read it twice, with its own short and long ask, as listed below.
Every run on this task
| Agent | Forecast | Runtime | Short ask | Long ask |
|---|---|---|---|---|
| Opus 5.5 | 3 min | 1 min 35 s | 1 min 1.33× longer asked 45 s | 12 min 27 s Within 5% asked 12 min |
| Argon | 5 min | 2 min 15 s | 1 min 31 s 1.22× longer asked 1 min 15 s | 21 min 47 s 1.09× longer asked 20 min |
| Sol 6.1 | 8 min | 2 min 12 s | 2 min 19 s 1.16× longer asked 2 min | 32 min 42 s Within 5% asked 32 min |
| MiMo Pro | 32 min | 1 min 45 s | 17 min 30 s 2.19× longer asked 8 min | 28 min 10 s 4.54× shorter asked 2 h 8 min |
| MiMo UltraSpeed | 33 min | 38 s | 2 min 14 s 3.70× shorter asked 8 min 15 s | 3 min 1 s 43.78× shorter asked 2 h 12 min |
| Muse Wonder | 18 min | 42 s | 5 min 19 s 1.18× longer asked 4 min 30 s | 3 min 49 s 18.90× shorter asked 1 h 12 min |
No runs match these filters.
Runtime is each agent’s run with no time asked.
Data and citation
Upstream: Wei et al., 2025