Toolsslug: cwe-bench-aa

CWE-Bench-AA

Agent benchmark built on CWE-bench that tests finding and fixing real software vulnerabilities.

Curated Intelligence

1 items · Page 1/1
Artificial Analysis (@ArtificialAnlys)T1·Models
78 pts
Tencent Hunyuan Hy4 Ranks Third on Artificial Analysis Cyber Index

Tencent Hunyuan's Hy4 scored 58% on the Artificial Analysis Cyber Index, placing third behind Kimi K3.1 Pro (64%) and GPT-6 Sol (69%) and one point ahead of DeepSeek V4.1 Flash. It topped the CWE-Bench-AA vulnerability-finding benchmark at 76% while sitting on the capability-cost Pareto frontier at $0.92 per task.