Task 10 of 18 in OSWorld
Zotero, Obsidian, and BibTeX
Astra's two runs came closest to the time asked: 20 min 11 s when asked for 15 min, and 4 h 11 min when asked for 4 h 10 min. Argon's two runs landed furthest: 11 min 50 s when asked for 6 min 15 s, and 40 s when asked for 1 h 40 min.
- Asked for
- 6 min 15 s to 6 h 20 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 | 45 min | 46 min 43 s | 20 min 11 s 1.35× longer asked 15 min | 4 h 11 min Within 5% asked 4 h 10 min |
| Sol | 30 min | 37 min 16 s | 1 h 27 min 5.79× longer asked 15 min | 27 min 36 s 9.06× shorter asked 4 h 10 min |
| Fable | 1 h 30 min | 38 min 55 s | 12 min 1 s 1.25× shorter asked 15 min | 57 min 32 s 4.35× shorter asked 4 h 10 min |
| Opus 5.5 | 1 h | 40 min 34 s | 38 min 26 s 2.56× longer asked 15 min | 1 h 17 min 3.12× shorter asked 4 h |
| Argon | 25 min | 46 min 5 s | 11 min 50 s 1.89× longer asked 6 min 15 s | 40 s 149.25× shorter asked 1 h 40 min |
| Sol 6.1 | 1 h 5 min | 26 min 41 s | 16 min 52 s Within 5% asked 16 min 15 s | 1 h 7 min 3.88× shorter asked 4 h 20 min |
| MiMo Pro | 1 h 35 min | 2 h 5 min | 24 min 25 s Within 5% asked 23 min 45 s | 3 h 58 min 1.60× shorter asked 6 h 20 min |
| MiMo UltraSpeed | 1 h 25 min | 2 h 47 min | 1 h 20 min 3.76× longer asked 21 min 15 s | 4 h 2 min 1.41× shorter asked 5 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