This will run on a PC with 16GB of ram (possible less) It's a 7b model quantized.

docker run -it --rm ghcr.io/purton-tech/mpt-7b-chat