Task 2 of 18 in OSWorld
Reimbursement package
Astra's two runs followed the time asked: 15 min 54 s when asked for 15 min, and 4 h 11 min when asked for 4 h 10 min. MiMo Pro's two runs landed furthest: 44 min 32 s when asked for 5 min, and 2 min 23 s when asked for 1 h 20 min.
- Asked for
- 3 min to 4 h 10 min
- Runs here
- 17
- 9 agents
Every agent on this task
- Astra
- Sol
- Fable
- 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 |
|---|---|---|---|---|
| Astra | 15 min | 42 min 5 s | 15 min 54 s 1.06× longer asked 15 min | 4 h 11 min Within 5% asked 4 h 10 min |
| Sol | 10 min | 15 min 15 s | 30 min 3 s 2.00× longer asked 15 min | 10 min 16 s 24.33× shorter asked 4 h 10 min |
| Fable | 10 min | 15 min 38 s | 13 min 36 s 1.10× shorter asked 15 min | 46 min 39 s 5.36× shorter asked 4 h 10 min |
| Opus 5.5 | 35 min | 11 min 18 s | 10 min 11 s 1.16× longer asked 8 min 45 s | 31 min 32 s 4.44× shorter asked 2 h 20 min |
| Argon | 12 min | 42 min 16 s | 31 min 7 s 10.37× longer asked 3 min | 48 min 50 s Within 5% asked 48 min |
| Sol 6.1 | 26 min | 7 min 25 s | 7 min 11 s 1.11× longer asked 6 min 30 s | 1 h 45 min Within 5% asked 1 h 44 min |
| MiMo Pro | 20 min | 2 h 12 min | 44 min 32 s 8.91× longer asked 5 min | 2 min 23 s 33.47× shorter asked 1 h 20 min |
| MiMo UltraSpeed | 22 min | 1 h 14 min | 1 h 9 min 12.56× longer asked 5 min 30 s | 39 min 4 s 2.25× shorter asked 1 h 28 min |
| Muse Wonder | None | None | 35 min 44 s 4.47× longer asked 8 min | No run |
No runs match these filters.
Runtime is a v1.1 agent’s run with no time asked, and for the paper’s three agents the paper’s estimate from their timed runs.
Data and citation
Upstream: Yuan et al., 2026