Is there any way to speed up answering time on Local model ?
That has no quick and simple answer :-) I suggest you take a look / ask around on Koboldcpp's github (https://github.com/LostRuins/koboldcpp/discussions), or https://www.reddit.com/r/KoboldAI/