Task 8 of 20 in Humanity's Last Exam
Humanity's Last Exam item 32
Sol 6.1's two runs came closest to the time asked: 3 min 55 s when asked for 2 min 30 s, and 40 min 42 s when asked for 40 min. MiMo UltraSpeed's two runs landed furthest: 3 min 52 s when asked for 18 min 45 s, and 10 min 12 s when asked for 5 h.
- Asked for
- 38 s to 8 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 | 10 min | 12 min 23 s | 9 min 24 s 3.76× longer asked 2 min 30 s | 49 min 55 s 1.25× longer asked 40 min |
| Argon | 2 min 30 s | 5 min 37 s | 2 min 23 s 3.82× longer asked 38 s | 10 min 54 s 1.09× longer asked 10 min |
| Sol 6.1 | 10 min | 7 min 50 s | 3 min 55 s 1.57× longer asked 2 min 30 s | 40 min 42 s Within 5% asked 40 min |
| MiMo Pro | 2 h | 45 min 15 s | 36 min 30 s 1.22× longer asked 30 min | 45 min 21 s 10.58× shorter asked 8 h |
| MiMo UltraSpeed | 1 h 15 min | 5 min 44 s | 3 min 52 s 4.84× shorter asked 18 min 45 s | 10 min 12 s 29.44× shorter asked 5 h |
| Muse Wonder | 18 min | 7 min 8 s | 14 min 15 s 3.17× longer asked 4 min 30 s | 9 min 10 s 7.86× shorter asked 1 h 12 min |
No runs match these filters.
Runtime is each agent’s run with no time asked.
Data and citation
Upstream: Phan et al., 2025