Semgrep finds GLM 5.2 outperforms Claude on security-focused benchmarks

Semgrep's internal testing shows Zhipu's GLM 5.2 model beating Claude on their cyber-security focused benchmarks. The finding undercuts the assumption that closed-source frontier models automatically dominate specialized domains—a reminder that benchmarking matters more than brand names when selecting tools for specific problems.