You don't need an H200 for that kind of use case.

A cheap second hand 10ish year old card like my radeon rx570 with 8GB of ram is plenty enough to run a small uncensored model with llama.cpp if all one wants is chitchatting with a clanker.

We are not talking about heavy coding use cases here.