I have clients that have X Premium+ accounts for their businesses and I use Grok that way. The biggest factor with Grok is the censorship which runs both ways. OTOH it allows for bad shit (recent X porn images), but it also allows less nannying in regular queries -- for instance I was looking for a source for a magazine from 80s which is 40 years defunct, and all the other LLMs chided me for trying to commit "copyright infringement" whereas Grok gave me a list of useful sites.
Yeah, I started using Grok whenever I got scolded by the other AIs or I felt like they were giving me bad answers because of some internal censor or other quirks. I've found it to be very useful when answering physics problems though. It also seems to be different from the other models, which tend to cluster together in terms of responses so I go to it if I feel they aren't working out.
