The quality of a chatbot has less to do with which model powers it and more to do with what it's connected to and what it's allowed to do.
Before starting a build, we ask clients three things: what source of truth should the bot answer from, what actions should it be able to take rather than just describe, and what should happen when it doesn't know the answer.
Get those three right, and the model choice becomes a much smaller decision than most teams expect.