But that also depends on your prompt. If you have it set that {{ char }} must speak in every response, that should happen more often or always (I mean you see the results in these tests here, always would be too good :D).

No you dont. But you need enough System Ram and its still very very slow like < 1tkn/s

Oh gott danke :).

Oh I was just thinking of asking if someone has done something like that yesterday. Thank you for the work! A Website about this would be look with consistent checks, but I guess its a lot of work.

Based on your knowledge, what is currently the best < 30B Roleplay Model? I prefer 20B for speed but that size doesnt seem to be trending :(