PetriBench - Easy: leaderboard
Metric: Exact-Match Accuracy (%). Source: arxiv.org. 36 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.8 Flash (High) | 96.8 |
| 2 | GPT-5.6 Sol (xHigh) | 96.4 |
| 3 | Grok 4.6 (xHigh) | 95.3 |
| 4 | Grok 4.6 (High) | 95.1 |
| 5 | Grok 4.6 (Medium) | 93.9 |
| 6 | GPT-5.6 Sol (High) | 92.9 |
| 7 | Claude Opus 5 (xHigh) | 92.7 |
| 8 | Gemini 3.1 Pro (Preview) (High) | 92.3 |
| 9 | Claude Opus 5 (High) | 90.4 |
| 10 | Grok 4.5 (High) | 88.2 |
| 11 | Gemini 3.8 Flash (Medium) | 86.6 |
| 12 | GPT-5.6 Terra (xHigh) | 85 |
| 13 | Claude Opus 5 (Medium) | 84.3 |
| 14 | Grok 4.5 (Medium) | 83.8 |
| 15 | GPT-5.6 Sol (Medium) | 82.7 |
Interactive version: theaggregate.ai/benchmark?slug=petribench-easy · How It Works · Data refreshed daily, snapshot 2026-09-20.