It felt like a rocket compared to GLM 5.2 though. Are Chinese models generally token-heavy?
https://artificialanalysis.ai/?cost=intelligence-vs-cost-per...