PetriBench - Deadlock (Hard): leaderboard
Exact-match accuracy on PetriBench's deadlock questions at the Hard structural-complexity level, one of six task families over Petri-net state spaces. Reasoning effort is part of the row identity. Higher is better. Source: arxiv.org/abs/2609.19883.
Metric: Exact-Match Accuracy (%). Source: arxiv.org. Saturation forecast: Around October 2027. 36 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol (xHigh) | 92 | #22 (GPT-5.6 Sol) |
| 2 | Claude Opus 5 (xHigh) | 82.5 | #16 (Claude Opus 5) |
| 3 | GPT-5.6 Sol (High) | 79 | #22 (GPT-5.6 Sol) |
| 4 | Claude Opus 5 (High) | 77.5 | #16 (Claude Opus 5) |
| 5 | GPT-5.6 Terra (xHigh) | 77 | #40 (GPT-5.6 Terra) |
| 6 | Claude Opus 5 (Medium) | 65 | #16 (Claude Opus 5) |
| 7 | GPT-5.6 Luna (xHigh) | 64 | #79 (GPT-5.6 Luna) |
| 8 | GPT-5.6 Terra (High) | 63.5 | #40 (GPT-5.6 Terra) |
| 9 | Grok 4.6 (xHigh) | 62 | #29 (Grok 4.6) |
| 10 | GPT-5.6 Luna (High) | 61 | #79 (GPT-5.6 Luna) |
| 11 | Grok 4.5 (Medium) | 59 | #54 (Grok 4.5) |
| 12 | Grok 4.5 (High) | 58.5 | #54 (Grok 4.5) |
| 13 | Gemini 3.8 Flash (High) | 55.5 | #30 (Gemini 3.8 Flash) |
| 14 | Gemini 3.1 Pro (Preview) (High) | 54.5 | #64 (Gemini 3.1 Pro (Preview)) |
| 15 | Grok 4.6 (High) | 54.5 | #29 (Grok 4.6) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=petribench-deadlock-hard · How It Works · Data refreshed daily, snapshot 2026-11-05.