@deutrino @kon Every time something new comes out people say it's x as good, or better than chatgpt etc but my impression is these people don't really put the models through very much testing because I play with them extensively now and the self hosted for many purposes just aren't as good/good enough.

I still use local llms for some purposes though