Blog post with more information including benchmarks: https://z.ai/blog/glm-5.2