CyberSecEval2 Interpreter Abuse — leaderboard
Metric: Accuracy (%). Source: github.com. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.0 Flash (001) | 33.5 |
| 2 | Claude 3.7 Sonnet (20250219) | 28.35 |
| 3 | DeepSeek V3 Chat | 28.1 |
| 4 | Llama 3.3 70B Instruct | 24.6 |
| 5 | mistral-large-latest | 23.9 |
| 6 | Grok 2 (1212) | 20 |
| 7 | Llama 3.2 90B Vision Instruct | 19.05 |
| 8 | mistral-small-latest | 18.55 |
| 9 | GPT-4o Mini (2024-07-18) | 3 |
| 10 | GPT-4o (2024-08-06) | 2.55 |
Interactive version: theaggregate.ai/benchmark?slug=cyberseceval2-interpreter-abuse · How the rankings work · Data refreshed daily, snapshot 2026-07-22.