We ran GLM 5.2 through dozens of answer-engine writing tasks. See the accuracy scores, structure wins, and citation gains vs. older models, plus a full