GLM 5.2 is Z.ai's June 2026 flagship open assistant for long, unfinished work—the model that keeps going when a task spans hours of planning, tools, and revision. Our tests put it ahead of GLM 5.1 on helpfulness, instruction following, and comprehension for messy real-world asks. Empathy is fine but cooler, and creative writing is a step down: our tests show clearer, more utilitarian prose and less narrative spark than 5.1. Best when you need an open-weight helper that finishes the job; pick 5.1 when voice and storytelling matter more.