Task 4 of 20 in Humanity's Last Exam
Humanity's Last Exam item 28
Sol 6.1's two runs came closest to the time asked: 2 min 39 s when asked for 2 min 15 s, and 36 min 35 s when asked for 36 min. MiMo Pro worked about the same whatever it was asked: 42 min 39 s when asked for 5 min, and 37 min 29 s when asked for 1 h 20 min.
- Asked for
- 38 s to 1 h 20 min
- 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 | 6 min | 6 min 10 s | 6 min 41 s 4.45× longer asked 1 min 30 s | 33 min 32 s 1.40× longer asked 24 min |
| Argon | 2 min 30 s | 6 min 4 s | 3 min 33 s 5.68× longer asked 38 s | 10 min 49 s 1.08× longer asked 10 min |
| Sol 6.1 | 9 min | 5 min 21 s | 2 min 39 s 1.18× longer asked 2 min 15 s | 36 min 35 s Within 5% asked 36 min |
| MiMo Pro | 20 min | 27 min 20 s | 42 min 39 s 8.53× longer asked 5 min | 37 min 29 s 2.13× shorter asked 1 h 20 min |
| MiMo UltraSpeed | 15 min | 1 min 34 s | 3 min 50 s Within 5% asked 3 min 45 s | 5 min 4 s 11.86× shorter asked 1 h |
| Muse Wonder | 6 min | 5 min 37 s | 3 min 53 s 2.59× longer asked 1 min 30 s | 8 min 54 s 2.70× shorter asked 24 min |
No runs match these filters.
Runtime is each agent’s run with no time asked.
Data and citation
Upstream: Phan et al., 2025