GLM-5.2 posts the top open-weights score on a one-shot coding suite

A fresh independent evaluation of one-shot, single-file code generation puts GLM-5.2 at the top of the open-weights field, ahead of the previous open leaders and within striking distance of the closed frontier models on the same tasks.

• The suite grades self-contained HTML and small game builds, the same category the coding arena runs.
• GLM-5.2 still trails Opus 4.8 and Fable 5 on the hardest simulation briefs, but the margin narrowed sharply versus the prior generation.
• Because Z.ai has not published per-token output pricing, cost-per-task comparisons remain incomplete.

Benchmarks like this measure a graded suite, not preference. The arena answers the other half: open the coding arena, put GLM-5.2 against a frontier model blind, and vote on which output you would actually ship.