I’m not talking about the speed I’m talking about the quality of output. I don’t think you understand how these models are transformed into “uncensored models” but a lot of the time using abliteration messed then up.