Vulnerability Detection on Security Review Benchmark 100-sample Baseline (C0 + SP0)
96.1C0 Detection RatePerplexity Sonar Pro
Evaluation Results
| Method | Links | |
|---|---|---|
| Perplexity Sonar ProClass=Commercial2026.02 | 96.1 | |
| Claude Opus 4.6Class=Commercial2026.02 | 95 | |
| GPT-5.2Class=Commercial2026.02 | 93 | |
| Gemini 2.5 ProClass=Commercial2026.02 | 91.9 | |
| DeepSeek ChatClass=Commercial2026.02 | 89 | |
| Llama 3.3 70BClass=Open-Source2026.02 | 72 | |
| Qwen 2.5 72BClass=Open-Source2026.02 | 70 | |
| DeepSeek-Coder-V2 16BClass=Open-Source2026.02 | 53 |