2 ms·with regular llama.cpp on a 3070ti I get 60tok/s TG with the 9B model, it's quite impressive.by ranger_danger 8mo agowith regular llama.cpp on a 3070ti I get 60tok/s TG with the 9B model, it's quite impressive.