Zhipu AI releases GLM-5.3: China's new open-weights model claims to be the strongest for code

Chinese AI startup Zhipu AI has released GLM-5.3, a model built on the same base as its predecessor GLM-5.2, with all improvements stemming from expanded post-training. The company claims GLM-5.3 is the most capable open-weights model for programming, with notable gains in agentic coding. A known weakness of previous leading Chinese models like Kimi or Qwen compared to US frontier models is cybersecurity. Zhipu AI therefore trained GLM-5.3 specifically with data and environments for detecting software vulnerabilities. The model reportedly developed the ability to reason through multi-step attack chains faster than expected. In collaboration with Chinese security teams, 2,436 vulnerabilities were found across 269 projects, some up to 40 years old, and these are documented in a public registry. GLM-5.3 is available now via the GLM Coding Plan and can be used in coding agents such as ZCode, Claude Code, or OpenCode. The model weights will be released as open source in two weeks once security assessments are completed.