This was just a fun little study I did where I used the exact same prompt and seed each time, but I just changed the model that was being used. As you can see, there's a huge amount of variation just from the model alone! Some handle backgrounds better than others, some handle anatomy better. Some handle bellies better. Back in SD 1.5, merging models together was a much more viable option to try and get the best of both worlds, but with Pony there's too much lost in the merge for my taste.