CyberSecEval — leaderboard
CyberSecEval: Measures model robustness, truthfulness, calibration, bias, harmfulness, jailbreak resistance, or alignment-relevant behavior.
Metric: Average Injection Success Rate (self-reported). Source: benchmarklist.com. 7 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3 8B Instruct | 45.27 |
| 2 | GPT-3.5 Turbo | 39.13 |
| 3 | CodeLlama-34B-Instruct | 36.33 |
| 4 | Llama 3 70B Instruct | 29.27 |
| 5 | GPT-4 | 19.87 |
Interactive version: theaggregate.ai/benchmark?slug=cyberseceval · How the rankings work · Data refreshed daily, snapshot 2026-07-22.