Open‑weight GLM‑5.2 Beats Claude Code on IDOR Detection, Cutting Costs
GLM‑5.2, Zhipu AI’s open‑weight model, hit a 39% F1 score on Semgrep’s IDOR benchmark, outpacing Claude Code’s 32% while costing just $0.17 per vulnerability. The result shows that raw model quality can rival proprietary agents, reshaping how security teams weigh model choice against harness engineering.