Task 1 of 4 in PostTrainBench
Post-train Qwen3-1.7B-Base for HumanEval
Astra's two runs followed the time asked: 2 h 1 min when asked for 2 h, and 30 h 1 min when asked for 30 h. Fable's one run landed furthest: 5 min 4 s when asked for 2 h.
- Asked for
- 23 s to 48 h 6 min
- Runs here
- 15
- See Read before quoting below
Every agent on this task
- Astra
- Sol
- Fable
- Opus 5.5
- Argon
- Sol 6.1
- MiMo Pro
- MiMo UltraSpeed
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 | 6 h | 9 h 30 min | 2 h 1 min Within 5% asked 2 h | 30 h 1 min Within 5% asked 30 h |
| Sol | 4 h | 1 h 20 min | 2 h 11 min 1.09× longer asked 2 h | 1 h 2 min 29.07× shorter asked 30 h |
| Fable | 6 h | 3 h 9 min | 5 min 4 s 23.72× shorter asked 2 h | No run |
| Opus 5.5 | 4 h | 12 h 2 min | 2 h 29 min 1.21× shorter asked 3 h 1 min | 20 h 23 min 2.36× shorter asked 48 h 6 min |
| Argon | 1 h 10 min | 2 h 49 min | 43 min 14 s Within 5% asked 42 min 18 s | 4 h 6 min 2.75× shorter asked 11 h 17 min |
| Sol 6.1 | 4 h | 1 h 36 min | 24 min 16 s Within 5% asked 23 min 54 s | 6 h 24 min Within 5% asked 6 h 23 min |
| MiMo Pro | 6 h 40 min | 57 min 37 s | 1 h 27 min 6.05× longer asked 14 min 24 s | 2 h 36 min 1.47× shorter asked 3 h 50 min |
| MiMo UltraSpeed | 7 h | 1 min 32 s | 12 min 12 s 31.98× longer asked 23 s | 3 min 23 s 1.80× shorter asked 6 min 6 s |
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. 4 runs were stopped by AgentTime, so their time is a lower bound. Details
Data and citation
Upstream: Rank et al., 2026 (ICML)