It looks like it takes three different models chat/ranking/ etc. What is the total cost for three API or VRAM requirements for reasonable local open source mode

It looks like it takes three different models chat/ranking/ etc. What is the total cost for three API or VRAM requirements for reasonable local open source models to run Continue?