Benchmark

ExploitGym

Exploit-development tasks.

Score history →

Model scores

#	Model	Score
1	GPT 5.6-Sol	33.7%
2	Claude Opus 5	22.1%
3	DeepSeek V4.1-Flash 0910	15.3%
4	GLM 5.3	15%
5	DeepSeek V4-Pro 0813	5.4%
6	DeepSeek V4-Flash 0731	1.8%