Task 15 of 18 in OSWorld
FreeCAD bracket
Astra's two runs came closest to the time asked: 19 min 38 s when asked for 15 min, and 4 h 11 min when asked for 4 h 10 min. Argon worked about the same whatever it was asked: 1 h 8 min when asked for 6 min 30 s, and 1 h 51 min when asked for 1 h 44 min.
- Asked for
- 6 min 30 s to 5 h
- Runs here
- 16
- See Read before quoting below
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 41 s | 19 min 38 s 1.31× longer asked 15 min | 4 h 11 min Within 5% asked 4 h 10 min |
| Sol | 35 min | 17 min 56 s | 16 min 23 s 1.09× longer asked 15 min | 35 min 40 s 7.01× shorter asked 4 h 10 min |
| Fable | 25 min | 50 min 46 s | 12 min 52 s 1.17× shorter asked 15 min | 8 h 30 min 2.04× longer asked 4 h 10 min |
| Opus 5.5 | 45 min | 2 h 36 min | 11 min 8 s Within 5% asked 11 min 15 s | 3 h 58 min 1.32× longer asked 3 h |
| Argon | 26 min | 1 h 41 min | 1 h 8 min 10.51× longer asked 6 min 30 s | 1 h 51 min 1.07× longer asked 1 h 44 min |
| Sol 6.1 | 55 min | 16 min 13 s | 14 min 42 s 1.07× longer asked 13 min 45 s | 41 min 52 s 5.26× shorter asked 3 h 40 min |
| MiMo Pro | 1 h 15 min | 4 h 4 min | 36 min 14 s 1.93× longer asked 18 min 45 s | 4 h 39 min 1.07× shorter asked 5 h |
| MiMo UltraSpeed | 45 min | 53 min 1 s | 12 min 22 s 1.10× longer asked 11 min 15 s | 1 h 49 min 1.66× shorter asked 3 h |
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.
Read before quoting. 1 run was stopped by AgentTime, so its time is a lower bound. Details
Data and citation
Upstream: Yuan et al., 2026