# They updated the image model! And to me it looks better than 2.5! — retained record

> Retained reddit record from reddit:ChatGPT.

Canonical: https://highsignal.app/data/records/a891cac913f3752a

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:ChatGPT)  published 2026-11-01 06:54 ingested 2026-11-02 02:37

[original source ↗](https://www.reddit.com/r/ChatGPT/comments/1xkwelm/they_updated_the_image_model_and_to_me_it_looks/)

## retained text



```
The title. I tried some of my prompts (basically, photorealistic images of people in city contexts) and damn it looks like they partly fixed the issues i pointed out in my [previous thread](https://www.reddit.com/r/ChatGPT/s/gG8bHjpkEz) about how bad image 2.5 was!

I'm so happy with that :)

for example i get much more adherence to the prompt, to the point that now using the same prompt i get TOO MUCH of what I've asked lol

but of course I'll need to adapt my prompts to the new model. let's start the exploration!

the only thing i noticed is that, using the same prompts, the overall tint is like washed out and muted, like the colours changed significantly. Hope it's still an issue of proper prompting and not a general problem of the model LMAO
```



document fetched 2026-11-02 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1xkwelm/they_updated_the_image_model_and_to_me_it_looks
