I have trained this LoRA for both SDXL and Z image Base and there is a clear jump in quality and prompt understanding between these models. 1 SDXL on these samples below and its easy to figure out which is the one. The variety of images that is possible after a training a tightly curated dataset seems limitless.
