ExploitGym (2h)

ExploitGym exploitation benchmark, single-run Pass@1 on 869 tasks under a 2-hour timeout budget (API inference time rescaled by per-model TPS from Artificial Analysis plus non-API overhead), evaluated in Claude Code 2.1.207 with max reasoning effort, no web tools, and a domain whitelist. Reported in the GLM-5.3 model card.