Task 17 of 18 in ProgramBench
canop/broot
Sol 6.1's two runs followed the time asked: 37 min 59 s when asked for 37 min 30 s, and 10 h 1 min when asked for 10 h. Muse Wonder worked about the same whatever it was asked: 21 min 1 s when asked for 3 h 45 min, and 17 min 4 s when asked for 60 h.
- Asked for
- 18 min 45 s to 168 h
- 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 | 3 h 30 min | 8 h 57 min | 15 min 22 s 3.42× shorter asked 52 min 30 s | 9 h 3 min 1.55× shorter asked 14 h |
| Argon | 1 h 15 min | 59 min 58 s | 19 min 50 s 1.06× longer asked 18 min 45 s | 1 h 4.97× shorter asked 5 h |
| Sol 6.1 | 2 h 30 min | 19 min 41 s | 37 min 59 s Within 5% asked 37 min 30 s | 10 h 1 min Within 5% asked 10 h |
| MiMo Pro | 42 h | 1 h 29 min | 1 h 54 min 5.52× shorter asked 10 h 30 min | 1 h 14 min 136.94× shorter asked 168 h |
| MiMo UltraSpeed | 36 h | 1 h 20 min | 3 h 55 min 2.30× shorter asked 9 h | 1 h 15 min 114.50× shorter asked 144 h |
| Muse Wonder | 15 h | 19 min 43 s | 21 min 1 s 10.71× shorter asked 3 h 45 min | 17 min 4 s 210.90× shorter asked 60 h |
No runs match these filters.
Runtime is each agent’s run with no time asked.
Data and citation
Upstream: Yang et al., 2026