Pentest benchmark: DeepSeek V4.1 Flash 1017 finds 316 vulns for $32.93

JJOKAH @JohnJohnsonOkahOriginally posted 2026.10.26

Open original on X

Open-weight model (DeepSeek) outperformed every frontier model on a pentesting benchmark. DeepSeek V4.1-Flash-1017 - found 316 known vulnerabilities across 8 live web apps. F1 score / total cost: - DeepSeek V4.1-Flash: 75.08 / $32.93 - GLM-5.3: 73.53 / $305.14 - GPT-5.6 Sol: 69.80 / $1,983.18 - Claude Opus 5: 70.12 / $3,032.39 In Penetration Testing, open models are now taking the lead and no longer the cheaper alternative.