GLM-5.2

GLM-5.2 is a Chinese large language model developed by Zhipu AI (智谱), positioned as the flagship coding model in the GLM series and widely regarded as the top domestic model for programming tasks . In June 2026, it scored 85.5 on the ZangAI benchmark, narrowly surpassing Claude Opus 4.8 (85.2) to take first place in a test measuring front-end and business-logic code generation . The model was praised for stable output quality without extreme failure cases, strong visual fidelity in generated interfaces, and performance that testers described as "completely on par with Opus 4.8" for one-shot web generation . It was still in preview access as of mid-June 2026, with the benchmark author noting they had to use an internal testing interface rather than a standard paid API .