This is a way to run open source models locally. You need the right hardware but it is a very efficient way to experiment with the newest models, fine tuning etc. ChatGPT uses massive model which are not practical to run on your own hardware. Privacy is also an issue for many people, particularly enterprise.