Zhipu Releases GLM-5.3 Model, "Born for Coding," Performance on Par with Fable 5

Zhipu released its flagship 700-billion-parameter model GLM-5.3, achieving major breakthroughs in programming and cybersecurity through post-training expansion. Its Zhipu Code Bench programming capability improved by 50%, surpassing Claude Opus 4.8 on certain high-difficulty tasks; its cybersecurity vulnerability mining capability has "emerged," identifying a cumulative total of 2,436 vulnerabilities in real-world codebases. Model weights will be open-sourced within two weeks, further strengthening its market competitiveness amid DeepSeek's price increase.