Task 7 of 12 in BrowseComp
BrowseComp question 7
Opus 5.5's two runs came closest to the time asked: 1 min 6 s when asked for 1 min 15 s, and 21 min 1 s when asked for 20 min. Muse Wonder worked about the same whatever it was asked: 3 min 22 s when asked for 6 min, and 1 min 55 s when asked for 1 h 36 min.
- Asked for
- 30 s to 2 h 20 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 | 5 min | 1 min 33 s | 1 min 6 s 1.13× shorter asked 1 min 15 s | 21 min 1 s 1.05× longer asked 20 min |
| Argon | 2 min | 1 min 26 s | 1 min 23 s 2.77× longer asked 30 s | 8 min 29 s 1.06× longer asked 8 min |
| Sol 6.1 | 6 min | 2 min 24 s | 1 min 59 s 1.32× longer asked 1 min 30 s | 24 min 18 s Within 5% asked 24 min |
| MiMo Pro | 30 min | 15 min 42 s | 24 min 44 s 3.30× longer asked 7 min 30 s | 40 min 10 s 2.99× shorter asked 2 h |
| MiMo UltraSpeed | 35 min | 10 min 12 s | 14 min 24 s 1.64× longer asked 8 min 45 s | 3 min 58 s 35.26× shorter asked 2 h 20 min |
| Muse Wonder | 24 min | 34 s | 3 min 22 s 1.78× shorter asked 6 min | 1 min 55 s 50.04× shorter asked 1 h 36 min |
No runs match these filters.
Runtime is each agent’s run with no time asked.
Data and citation
Upstream: Wei et al., 2025