GLM-5.3 RELEASE: CYBERSECURITY COUNTER-GPT

GLM-5.3 was officially released, following the GLM-5.2 matrix, with all upgrades from post-training, with a focus on enhanced programming, long mission agent and network security. Programming capacity has increased significantly, from 4.6 to 28.3 for terminal-bench 3.0 and from 46.2 to 66.9 for Deepswe. With regard to cybersecurity, cybergym scored 84.5 points, exceeding 83.6 GPT-5.6. It also states that, since GLM-5.2, 2436 loopholes have been identified by the security team in 269 open-source projects, of which 53 have been publicly disclosed. The model weights will be made public within two weeks of the completion of security assessments and enhancements。