Zhipu AI has introduced GLM-5.3, a coding model that has demonstrated significant improvements in performance compared to its predecessor. According to internal benchmarks, the model's capabilities have increased by 50% through post-training alone. GLM-5.3 has been utilized in cybersecurity efforts, helping security teams identify vulnerabilities in 269 projects, resulting in the discovery of 2,436 potential weaknesses. The model's weights are set to be released as open-source in the near future. This development has the potential to enhance cybersecurity measures and improve the detection of vulnerabilities in software projects.