Running locally is even worse for this, because if you're running 4 jobs at once they just run at 1/4 speed. Not literally, you can make up some of the difference with batching, but you have limited resources instead of spreading your requests out on an API provider's nodes.