Chinese AI lab releases GLM-5.2, tops independent benchmarks

What happened

Z.ai released GLM-5.2, a 753-billion-parameter text-only model, under an MIT license on June 16th. The model ranks #1 on the Artificial Analysis Intelligence Index v4.1 with a score of 51, ahead of MiniMax-M3 (44), DeepSeek V4 Pro (44), and Kimi K2.6 (43). It also ranks #2 on the Code Arena WebDev leaderboard, behind only Claude Fable 5.

Why it matters

GLM-5.2 achieves top-tier performance on major independent benchmarks while remaining available as open-weights software under a permissive license. This means researchers and businesses can deploy and modify it without licensing restrictions. The model's 1 million token context window (up from GLM-5.1's 200,000) allows it to process and work with substantially longer documents or conversations.

What to watch

The model generates substantially more output tokens per task than competitors—43k tokens per benchmark task, compared with 26k for GLM-5.1 and 24k for MiniMax-M3. On OpenRouter, input costs $1.40/million tokens and output costs $4.40/million tokens, making it cheaper than GPT-5.5 ($5/$30) and Claude Opus 4.5-4.8 ($5/$25).