TLDR Tech / InfoSec
Semgrep: GLM 5.2 Beats Claude in Cybersecurity Coding Benchmarks
GLM 5.2, an open-weight model from Zhipu AI, achieved a 39% F1 score in Semgrep's IDOR detection benchmark, outperforming Claude Code's 32% at a cost of $0.17 per true finding, though Semgrep's own multimodal pipeline with endpoint scaffolding reached 53% to 61%, underscoring that harness design significantly influences real-world effectiveness beyond raw model capability. The results highlight GLM 5.2 as a cost-efficient, competitive option for security-focused code analysis, particularly when deployed within optimized evaluation frameworks.
Read the full story