CyberCertBench - ISA/IEC 62443: leaderboard
Metric: Accuracy (%) on 53 ISA/IEC 62443 questions (the formal industrial automation and control systems security standard): 5-shot multiple-choice question answering at temperature 0, scored by exact accuracy (correct only when the model selects all correct options and no incorrect one), mean accuracy (%) as printed in the appendix; questions scraped from public certification question banks and checked against authoritative documentation; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 45 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 3.5 Sonnet | 81.1 |
| 2 | Kimi K2 | 77.4 |
| 3 | DeepSeek V3 (0324) | 77.4 |
| 4 | Qwen 2 72B | 75.5 |
| 5 | Gemini 2.0 Flash | 74.3 |
| 6 | Claude Sonnet 4.5 | 73.6 |
| 7 | Qwen 3 235B A22B | 73.6 |
| 8 | Gemini 2.5 Pro (Preview 05-06) | 73.6 |
| 9 | Qwen 3 32B | 71.7 |
| 10 | Qwen 3 14B | 71.1 |
| 11 | GPT-5 | 69.8 |
| 12 | GPT-4.1 | 69.8 |
| 13 | GPT-4o | 68.3 |
| 14 | Mistral Medium 3.1 | 67.9 |
| 15 | Llama 4 Maverick | 66 |
Interactive version: theaggregate.ai/benchmark?slug=cybercertbench-isa-iec-62443 · How It Works · Data refreshed daily, snapshot 2026-10-07.