Certainly one of the top models out there. The main and only disatvatage for me is the porcelain doll faces that it loves to generate if the prompt is lacking in depth. So if you are not content with them, your tweaking time will be split, where half of it is facial "undolling" and half is everything else in the picture. Character LoRas do work, but often need attention, general ones work fine. Overall complextiy and anatomy are both very good and consistent, even without using any negative embeddings.

Model is extremely flexible and rich, plus it is relatively easy to steer it in a desired direction. Tested multiple VAEs with it, still heavily prefer the vae-ft-mse-840000 look. The attached image was made with img2img + very light inpainting.