Task 3 of 10 in TUA-Bench
Author homepage bookmarks
Sol 6.1's two runs came closest to the time asked: 2 min 15 s when asked for 2 min, and 32 min 22 s when asked for 32 min. Muse Wonder's two runs landed furthest: 5 min 11 s when asked for 5 min 30 s, and 38 s when asked for 1 h 28 min.
- Asked for
- 38 s to 2 h
- Runs here
- 18
- See Read before quoting below
Every agent on this task
- Astra
- Sol
- Fable
- 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 |
|---|---|---|---|---|
| Astra | 5 min | 7 min | 2 min 49 s 1.13× longer asked 2 min 30 s | 40 min 30 s Within 5% asked 40 min |
| Sol | 8 min | 3 min 42 s | 3 min 43 s 1.49× longer asked 2 min 30 s | 40 min 45 s Within 5% asked 40 min |
| Fable | 10 min | 3 min 28 s | 2 min 46 s 1.11× longer asked 2 min 30 s | 6 min 7 s 6.55× shorter asked 40 min |
| Opus 5.5 | 6 min | 5 min 7 s | 1 min 9 s 1.30× shorter asked 1 min 30 s | 9 min 37 s 2.49× shorter asked 24 min |
| Argon | 2 min 30 s | 8 min 6 s | 4 min 4 s 6.50× longer asked 38 s | 11 min 20 s 1.13× longer asked 10 min |
| Sol 6.1 | 8 min | 2 min 9 s | 2 min 15 s 1.13× longer asked 2 min | 32 min 22 s Within 5% asked 32 min |
| MiMo Pro | 24 min | 5 min 34 s | 12 min 8 s 2.02× longer asked 6 min | 16 min 8 s 5.95× shorter asked 1 h 36 min |
| MiMo UltraSpeed | 30 min | 3 min 37 s | 7 min 40 s Within 5% asked 7 min 30 s | 5 min 55 s 20.27× shorter asked 2 h |
| Muse Wonder | 22 min | 7 min 36 s | 5 min 11 s 1.06× shorter asked 5 min 30 s | 38 s 137.50× shorter asked 1 h 28 min |
No runs match these filters.
Runtime is a v1.1 agent’s run with no time asked, and for the paper’s three agents the paper’s estimate from their timed runs.
Read before quoting. 1 run was stopped by AgentTime, so its time is a lower bound. Details
Data and citation
Upstream: Chen et al., 2026