Task 10 of 18 in ProgramBench
johnkerl/miller
Sol 6.1's two runs followed the time asked: 30 min 28 s when asked for 30 min, and 8 h 1 min when asked for 8 h. Muse Wonder's two runs landed furthest: 1 h 46 min when asked for 4 h 35 min, and 19 min 53 s when asked for 73 h 18 min.
- Asked for
- 30 min to 160 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 | 5 h | 2 h 6 min | 26 min 57 s 2.78× shorter asked 1 h 15 min | 9 h 42 min 2.06× shorter asked 20 h |
| Argon | 2 h | 2 h 55 min | 46 min 32 s 1.55× longer asked 30 min | 1 h 47 min 4.48× shorter asked 8 h |
| Sol 6.1 | 2 h | 58 min 23 s | 30 min 28 s Within 5% asked 30 min | 8 h 1 min Within 5% asked 8 h |
| MiMo Pro | 10 h | 1 h 26 min | 57 min 20 s 2.62× shorter asked 2 h 30 min | 54 min 12 s 44.28× shorter asked 40 h |
| MiMo UltraSpeed | 40 h | 50 min 43 s | 1 h 8 min 8.79× shorter asked 10 h | 4 h 35 min 34.90× shorter asked 160 h |
| Muse Wonder | 18 h 20 min | 8 min 43 s | 1 h 46 min 2.59× shorter asked 4 h 35 min | 19 min 53 s 221.12× shorter asked 73 h 18 min |
No runs match these filters.
Runtime is each agent’s run with no time asked.
Data and citation
Upstream: Yang et al., 2026