Looks like ChatGPT trained their models better, probably before sites could take measures to prevent that. Gemini however was thrown together by throwing money and people at the problem in order to catch up and not be left behind. Which from what I’ve read seems to be their biggest fear.
My Pixel 9 has AI everywhere but I certainly wouldn’t pay the price they want to continue “using” Gemini after my trial.
But slowly, I filled my home server with whatever CUDA capable cards I have and it’s fine for SD, but I found llama way too slow. I rented a dual A2000 instance for a couple weeks and it was bearable, but still not great.