Semgrep

Semgrep: GLM 5.2 outperforms Claude on cybersecurity benchmarks

Semgrep published a blog post reporting that GLM 5.2 beats Claude on their internal cybersecurity benchmarks, framed as a 'we have Mythos at home' comparison. The post appears to evaluate models on cyber-specific tasks relevant to Semgrep's code security tooling. This is a practitioner-level benchmark comparison from a security-focused company, providing real-world signal on model performance in a specialized domain.