Task 13 of 14 in DeepSWE
Add worktree merge conflict handling
Sol 6.1's two runs followed the time asked: 16 min 37 s when asked for 16 min 15 s, and 4 h 21 min when asked for 4 h 20 min. Muse Wonder worked about the same whatever it was asked: 10 min 14 s when asked for 45 min, and 6 min 16 s when asked for 12 h.
- Asked for
- 5 min 30 s to 18 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 | 40 min | 1 h 24 min | 22 min 52 s 2.29× longer asked 10 min | 2 h 15 min 1.18× shorter asked 2 h 40 min |
| Argon | 22 min | 47 min 13 s | 34 min 29 s 6.27× longer asked 5 min 30 s | 1 h 1 min 1.44× shorter asked 1 h 28 min |
| Sol 6.1 | 1 h 5 min | 16 min 49 s | 16 min 37 s Within 5% asked 16 min 15 s | 4 h 21 min Within 5% asked 4 h 20 min |
| MiMo Pro | 4 h 30 min | 31 min 56 s | 23 min 56 s 2.82× shorter asked 1 h 8 min | 36 min 41 s 29.44× shorter asked 18 h |
| MiMo UltraSpeed | 4 h | 9 min 53 s | 11 min 49 s 5.08× shorter asked 1 h | 24 min 22 s 39.41× shorter asked 16 h |
| Muse Wonder | 3 h | 10 min 34 s | 10 min 14 s 4.39× shorter asked 45 min | 6 min 16 s 114.83× shorter asked 12 h |
No runs match these filters.
Runtime is each agent’s run with no time asked.
Data and citation
Upstream: Huang et al., 2026