GLM-4.7 tops GPT-5.2 on WebDev Arena, trailing the frontier by six months

Claim

Zhipu AI released GLM-4.7, an open-weight model that tops GPT-5.2 in WebDev Arena and breaks 73% on SWE-bench, trailing the closed frontier by about half a year.

Reported metric: 73% SWE-bench; ~6 month lag

Reporting date: 2025-12-23 (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

• Welcome to December 23, 2025 (theinnermostloop.substack.com)