GLM-5.3 leads all open models at Elo 1770 with a 19% cheaper cost per task

GLM-5.3, by China’s Z.ai, is now #1 in the world for open weights. It scores 60 on the top benchmark (tied with Kimi K3), achieves an Elo of 1770 in agentic tasks (second behind Claude Opus 5 at 1855), and costs $0.68/task (19% cheaper than next best).

The open weights release was pushed back 2 weeks because GLM-5.3 was too good at discovering flaws (a kind of “cyberweapon” that can’t be shared publicly). Now there’s no question that Chinese open models are viable — they’re the best when you need independent thought, not just question answering.