To counter this, I tried Openrouter's chat interface, but that is painfully slow. I'm trying now Gemini 2.5 in Google AI studio and it is also slow.
What is the underlying reason for this? I understand the backend takes a lot of computation, but frontend?