There are free and offline options, like Llama.cpp, but you will have to pay by giving up your privacy to Meta (or similar large companies)
Also, running models locally requires good hardware to get acceptable performance. It's still a large barrier to entry.