Semgrep Benchmarks GLM-5.2 Against Claude, Finds Higher IDOR F1

Semgrep researchers recently benchmarked the GLM-5.2 model against Anthropic's Claude, finding the former achieved a higher F1 score in identifying insecure direct object reference vulnerabilities. This comparison highlights an increasing trend of specialized, smaller models demonstrating competitive performance in automated security testing compared to larger, general-purpose LLMs.