Task 12 of 18 in ProgramBench
typst/typst
Sol 6.1's two runs followed the time asked: 26 min 37 s when asked for 26 min 15 s, and 7 h when asked for 7 h. Muse Wonder's two runs landed furthest: 10 min 38 s when asked for 7 h 30 min, and 22 min 37 s when asked for 120 h.
- Asked for
- 26 min 15 s to 120 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 | 4 h | 10 h 25 min | 17 min 34 s 3.42× shorter asked 1 h | 15 h 41 min Within 5% asked 16 h |
| Argon | 1 h 50 min | 1 h 42 min | 7 min 22 s 3.73× shorter asked 27 min 30 s | 1 h 23 min 5.29× shorter asked 7 h 20 min |
| Sol 6.1 | 1 h 45 min | 53 min 13 s | 26 min 37 s Within 5% asked 26 min 15 s | 7 h Within 5% asked 7 h |
| MiMo Pro | 21 h | 1 h 14 min | 2 h 25 min 2.18× shorter asked 5 h 15 min | 36 h 22 min 2.31× shorter asked 84 h |
| MiMo UltraSpeed | 20 h | 1 h 7 min | 1 h 13 min 4.12× shorter asked 5 h | 1 h 22 min 58.54× shorter asked 80 h |
| Muse Wonder | 30 h | 15 min 52 s | 10 min 38 s 42.33× shorter asked 7 h 30 min | 22 min 37 s 318.30× shorter asked 120 h |
No runs match these filters.
Runtime is each agent’s run with no time asked.
Data and citation
Upstream: Yang et al., 2026