Bard hasn’t been using Google’s best language models. I believe it just got an upgrade, however, and I’m now getting output that is significantly more coherent and useful than ChatGPT’s. It’s also a helluva lot faster, though that could owe to the limited access.
I mean they’ve literally said they are using a much smaller parameter model at first to while testing. you can not believe it if you want but that’s the very obvious answer.