Last week, GLM 5.3 flash was the most used model in the world (they made it free for that week). When the week ended, they announced that it had been a stress test for their entirely chinese inference infrastructure.

Goodbye Nvidia.