@frap129
@lemmy.maples.devI use koboldcpp with the vicuna model. Reasonably fast generation (<1 minute) on a 4th gen i7, would probably be on par with chatgpt in terms of speed if you used a GPU.
Oddly enough, Active has been the best for me. When I sort by Hot, I get almost entirely 2+ year old posts. Guess it all depends what communities you've joined
I joined Lemmy a few weeks ago when a lot of other reddit refugees were making the switch. A lot of the servers were overloaded, so I just went ahead and hosted my own. My instance is never overloaded because I'm the only user, and because Lemmy is federated I can subscribe to communities on any instance I want.