yes this was our first launch and we continue to improve our models. we have already made an update after this launch, now you can try it immediately and get much better results.
No clear benefit of using this product. Might as well use Dalle2/Midjourney/Stable Diffusion directly.
Edit: You guys can improve my training a version of SD specifically trained for creating logos. I believe the current model you are using is very less restricted which results in logo not being up to the mark.