Woofun AI reports that Silkbrain (02513.HK) released the open-weight model GLM-5.2, which researchers claim rivals Anthropic's Mythos in vulnerability detection. Semgrep data indicates GLM-5.2 outperformed Claude Opus 4.8 in specific benchmarks, with both models matching Mythos under further instruction.

However, its availability on standard hardware increases potential exploitation risks.