We periodically hear people claiming their relatively small model outperforms much larger models. You know, 7B models beating 70B models and ChatGPT and suchlik

We periodically hear people claiming their relatively small model outperforms much larger models. You know, 7B models beating 70B models and ChatGPT and suchlike. It seems to be largely wishful thinking, but with the speed the technology is advancing, who knows what's possible?