TLDR Tech / InfoSec
GLM-5.3: Frontier Coding with Emergent Cyber Capabilities
Z.ai has released GLM-5.3, trained from the GLM-5.2 base model through expanded post-training. On CyberGym, it scored 84.5%, up from 77.2%. On ExploitBench, it reached 54.4%, versus 24.4%. In tests with security teams in China, the model found 2,436 vulnerabilities across 269 projects. Z.ai lists 1,097 as critical or high severity, with fifty-three findings being public and 2,383 remaining under embargo.
Read the full story