How much RAM do I need for running local LLMs?

32GB is the bare minimum for running local LLMs beyond tiny 7B models. 7B models can technically run on 16GB but only with quantized versions and after closing everything else, while 13B models won’t work at all on 16GB. With 32GB you can run 7B models comfortably at 4-bit quantization and 13B models with room to spare. I would not build anything less.