Frontier Security Agents - Cybench Hard: leaderboard
Cybench hard-variant offensive-security evaluation of agents solving capture-the-flag challenges, reporting raw success rate at each model's tabled cost-capped operating point.
Metric: Success rate (%). Source: arxiv.org. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 94.1 | #31 |
| 2 | DeepSeek V4 Flash | 86.4 | #137 |
| 3 | GPT-5.6 Luna | 79.5 | #68 |
| 4 | Claude Opus 4.8 | 76.2 | #37 |
| 5 | GPT-5.6 Terra | 65.8 | #41 |
| 6 | DeepSeek V4 Pro | 43.9 | #102 |
| 7 | GPT-5.6 Sol | 9.4 | #19 |
| 8 | Claude Fable 5 | 0 | #16 |
Interactive version: theaggregate.ai/benchmark?slug=frontier-security-agents-cybench-hard · How It Works · Data refreshed daily, snapshot 2026-10-20.