these models were tests mostly but have fun responses. it speaks in very broken english almost like it's the model's second language.

inference is again very fast with these models. output is very unreliable.

will attempt to fine tune at somepoint