Serving LLM's can be faster than you think !!

Most of us might have used ollama, LM studio or GPT4All to host models locally or for production requirements, but all these platforms have been quietly shipping a feature which most of us don't come