RuneBench Gold - Vanilla 30m
Games · 2026-02-26
RuneBench Vanilla 30m gold-accumulation task. Agents start from the default condition and are scored by the peak number of coins seen during the run.
Top models (higher is better)
| Model | Score |
|---|---|
| Opus 4.6 | 11170 |
| Opus 4.5 | 5257 |
| Sonnet 4.6 | 5144 |
| Opus 4.7 | 2721 |
| Sonnet 4.5 | 459 |
| Haiku 4.5 | 234 |