It will run faster if your local machine is faster than the cloud, but otherwise it's the same. I don't think I tried starting CI from my machine, per se, I would instead point codeberg to my machine, push to codeberg, and codeberg would send the job to my local runner. You probably can do it entirely locally.

The main benefit for me was that I can run CUDA workflows without needing to pay extra for GPUs. I've not really thought about what happens if you want to deploy to an architecture you don't own. I can't recall what architectures they provide.

Note this is with woodpecker CI. I think they're migrating to a new CI at some point.