Good joke, thanks, but I will explain anyway: to me 30 billion parameters or smaller is small since I can run it using Ollama on my home computer. I managed a deep learning team at Capital One and our ‘large’ models were less than 20 million parameters. :-)