# After initial tests, Qwen Image 2.1 quality is below that of GPT Image 1.0. — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/b5ef9f476e351884

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-09-20 13:44 ingested 2026-09-21 02:37

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1wlhh8v/after_initial_tests_qwen_image_21_quality_is/)

## retained text



```
https://preview.redd.it/k1v42b2xloqh1.jpg?width=2048&format=pjpg&auto=webp&s=fcf209d9ddfe5de42ec4f0c1192cf400a5b5de57

So, I’m not impressed. Based on the workflow available in Comfy, the results look like the mentally challenged sibling of GPT Image 1.0.

There are a lot of errors and artifacts in the generated images.

On the plus side, it’s reasonably fast—generating a 4K image on an RTX 5080 takes 90 seconds.

It supports kinda adult content, but the anatomy is messed up.

Fingers are wrong, faces in picturs on the wall are wrong, books on the desk, etc.

 

Added Krea 2 result, same prompt:

https://preview.redd.it/rp1cfi71toqh1.jpg?width=1448&format=pjpg&auto=webp&s=7c61ffb2af711e557c424c0308c6d6e55d39321a
```



document fetched 2026-09-21 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1wlhh8v/after_initial_tests_qwen_image_21_quality_is
