What "solution" am I using when I choose "use in Diffusers"?

I was about to use a model with an Inference Endpoint, but noticed that there is the option of using it with Diffusers which is 1) very comfortable and 2) free (?).

This is the model I’m talking about: https://huggingface.co/eykarim/stable-diffusion-v1/

I’ve also noticed that when I use Diffusers, it’s quite resource-heavy on the machine I’m “using” it on. My original understanding was that the computation would be done on HuggingFace’s servers when I do this:

pipeline = DiffusionPipeline.from_pretrained("eykarim/stable-diffusion-v1", use_auth_token=True)

But is that really the case? What is the “product/solution” behind Diffusers? What is the relation between Diffusers and the paid Inference Endpoints? Guess I need a little orientation here…