Z.ai released full open weights for GLM-5.2 on June 16, a text-only mixture-of-experts model with 753 billion total parameters under an MIT license, after giving coding-plan subscribers access on June 13 (Simon Willison). The model carries a 1 million token context window, up from 200,000 in GLM-5.1. Artificial Analysis places it as the leading open-weight model on its Intelligence Index v4.1 with a score of 51, ahead of MiniMax-M3 and DeepSeek V4 Pro at 44 each, and it ranks second on the Code Arena WebDev leaderboard for frontend work behind Claude Fable 5. Pricing across most OpenRouter providers runs about $1.40 per million input tokens and $4.40 per million output tokens, though the model is notably token-hungry, using roughly 43,000 output tokens per task against GLM-5.1’s 26,000 (Simon Willison).
Separately, Willison highlighted Charity Majors arguing that 2025 turned the economics of code production upside down: where generating code was once slow and expensive, it became effectively free and instant, shifting lines of code from something treasured, reused and carefully curated to something disposable and regenerable, a change she frames as demanding more engineering discipline rather than less (Simon Willison). He also published a click-to-play web component, a progressive-enhancement pattern that lazily loads GIFs on demand (Simon Willison).