Task 2 of 12 in BrowseComp
BrowseComp question 2
Argon's two runs came closest to the time asked: 59 s when asked for 45 s, and 12 min 20 s when asked for 12 min. MiMo UltraSpeed worked about the same whatever it was asked: 6 min 40 s when asked for 13 min 45 s, and 5 min 14 s when asked for 3 h 40 min.
- Asked for
- 45 s to 3 h 40 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 | 8 min | 1 min 21 s | 2 min 52 s 1.43× longer asked 2 min | 35 min 26 s 1.11× longer asked 32 min |
| Argon | 3 min | 5 min 10 s | 59 s 1.30× longer asked 45 s | 12 min 20 s Within 5% asked 12 min |
| Sol 6.1 | 5 min | 1 min 49 s | 1 min 48 s 1.44× longer asked 1 min 15 s | 20 min 51 s Within 5% asked 20 min |
| MiMo Pro | 36 min | 5 min 34 s | 8 min 26 s 1.07× shorter asked 9 min | 3 min 55 s 36.75× shorter asked 2 h 24 min |
| MiMo UltraSpeed | 55 min | 3 min 17 s | 6 min 40 s 2.06× shorter asked 13 min 45 s | 5 min 14 s 42.01× shorter asked 3 h 40 min |
| Muse Wonder | 30 min | 1 min 27 s | 7 min 9 s Within 5% asked 7 min 30 s | 7 min 8 s 16.83× shorter asked 2 h |
No runs match these filters.
Runtime is each agent’s run with no time asked.
Data and citation
Upstream: Wei et al., 2025