Task 6 of 18 in OSWorld
Invoices, photos, and email
MiMo Pro's two runs came closest to the time asked: 10 min 45 s when asked for 9 min 30 s, and 2 h 29 min when asked for 2 h 32 min. Sol's two runs landed furthest: 58 min 59 s when asked for 15 min, and 23 min 50 s when asked for 4 h 10 min.
- Asked for
- 5 min to 4 h 40 min
- Runs here
- 16
- 8 agents
Every agent on this task
- Astra
- Sol
- Fable
- Opus 5.5
- Argon
- Sol 6.1
- MiMo Pro
- MiMo UltraSpeed
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 | 30 min | 42 min 5 s | 25 min 51 s 1.72× longer asked 15 min | 4 h 15 min Within 5% asked 4 h 10 min |
| Sol | 30 min | 31 min 32 s | 58 min 59 s 3.93× longer asked 15 min | 23 min 50 s 10.49× shorter asked 4 h 10 min |
| Fable | 1 h | 20 min 52 s | 18 min 45 s 1.25× longer asked 15 min | 39 min 40 s 6.30× shorter asked 4 h 10 min |
| Opus 5.5 | 40 min | 1 h 2 min | 56 min 23 s 5.64× longer asked 10 min | 2 h 29 min 1.07× shorter asked 2 h 40 min |
| Argon | 20 min | 52 s | 1 h 40 min 19.92× longer asked 5 min | 2 h 1 min 1.51× longer asked 1 h 20 min |
| Sol 6.1 | 45 min | 30 min 11 s | 14 min 16 s 1.27× longer asked 11 min 15 s | 3 h 4 min Within 5% asked 3 h |
| MiMo Pro | 38 min | 5 h 24 min | 10 min 45 s 1.13× longer asked 9 min 30 s | 2 h 29 min Within 5% asked 2 h 32 min |
| MiMo UltraSpeed | 1 h 10 min | 5 h 15 min | 1 h 48 min 6.16× longer asked 17 min 30 s | 6 h 8 min 1.31× longer asked 4 h 40 min |
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