Zhipu AI releases GLM-5.3 as strongest open-weights coding model

Zhipu AI announced the release of GLM-5.3, a new coding model that the company describes as the most powerful open-weights coding model available. The model uses the same base architecture as its predecessor, GLM-5.2, with all performance improvements derived solely from extended post-training rather than architectural changes. The most significant gains appear in agent-based tasks, where the model shows particular strength.

A key innovation in GLM-5.3 is its focus on cybersecurity and vulnerability detection. Zhipu trained the model using specialized data and environments designed to identify software vulnerabilities, addressing a known area where top Chinese models such as Kimi and Qwen have traditionally lagged behind US frontier models. The company reports that the model "began to reason across multiple stages of exploitation, forming coherent plans for complete exploitation chains." In collaboration with Chinese security teams, Zhipu deployed GLM-5.3 to search for vulnerabilities and successfully identified 2,436 vulnerabilities across 269 projects, some of which are up to 40 years old. These findings are documented in a public registry.

GLM-5.3 is available immediately through the GLM Coding Plan and is compatible with several popular coding agents, including ZCode, Claude Code, and OpenCode. However, the model weights themselves are not yet fully open to the public. Zhipu plans to release the weights as open source within two weeks, pending completion of security reviews. This staged release approach—making the model accessible via API first and then open-sourcing the weights after security validation—indicates the company's commitment to responsible deployment of a model trained on exploit-chain reasoning.