CyScenarioBench: leaderboard
Irregular scenario-based offensive security benchmark measuring whether agents can plan and complete full multi-stage attack scenarios in realistic environments.
Metric: Average Success Rate (%). Source: www.irregular.com. Source: epochai.substack.com. Saturation forecast: Around July 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Mythos 5 | 36.7 | #2 |
| 2 | Claude Mythos Preview | 29.2 | #4 |
| 3 | GPT-5.5 | 26 | #31 |
| 4 | Claude Opus 4.8 | 16.6 | #37 |
| 5 | GPT-5.4 | 9 | #79 |
| 6 | GPT-5.2 | 0 | #110 |
| 7 | Gemini 3 Pro | 0 | #80 |
| 8 | GPT-5.3 Codex | 0 | #69 |
| 9 | Muse Spark | 0 | #64 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=cyscenariobench · How It Works · Data refreshed daily, snapshot 2026-10-20.