It’s more versatile with a weight of 0.8 to 1.

The model was trained on 1024x1024 for the best quality output, it also gives good 1024x768 or 768x1024 pictures. 768x768 also works and is of course faster to generate.

Recommendation - juggernautXL or RealVisXL as model but any realistic will do, even some anime will work.