Chinese AI lab Z.ai released GLM-5.2, an open-weights text model that ranks as the leading open-source AI on independent benchmarks, though it consumes significantly more tokens than competitors.

What happened: Z.ai released GLM-5.2 on June 16th under an MIT license. The 753-billion-parameter model has a 1 million token context window, up from GLM-5.1's 200,000. On the Artificial Analysis Intelligence Index v4.1, GLM-5.2 scored 51, ranking it above MiniMax-M3 (44), DeepSeek V4 Pro (44), and Kimi K2.6 (43). It also ranks 2nd on the Code Arena WebDev leaderboard for front-end web development tasks, behind only Claude Fable 5.

Why it matters: GLM-5.2 is now the leading open-weights model on a widely respected independent benchmark, meaning businesses using open-source AI can now access a top-tier option without relying on proprietary systems. However, the model generates substantially more output tokens per task—43k tokens on Intelligence Index tasks—compared to GLM-5.1 (26k), DeepSeek V4 Pro max (37k), and MiniMax-M3 (24k), which may increase inference costs.

What to watch: GLM-5.2 is available via OpenRouter from 9 providers at $1.40/million for input and $4.40/million for output tokens—less expensive than GPT-5.5 ($5/$30) or Claude Opus 4.5-4.8 ($5/$25).