Zhipu releases GLM-5.1-highspeed, claiming a large-model API speed record

智谱GLM-5.1高速版发布：刷新全球大模型API速度纪录

Zhipu released the GLM-5.1-highspeed API to selected enterprise customers on May 22, with a claimed output speed of 400 tokens/s, built by the GLM team and TileRT team through system-level optimization.

Why it matters: HKR-H/K/R all pass: Zhipu’s GLM-5.1 high-speed API has a concrete 400 tokens/s claim and domestic flagship-model relevance. Test setup, pricing, and availability are not disclosed, so it stays in the 78–84 band.