Do the “Projects” feature on the ChatGPT web app and the “Projects” feature in the desktop app use different image generation models?
Heat trend
The percentage is based on available heat signal, not comment count or independent people.
OpenAI model activity is surfacing — worth tracking for capability changes, ecosystem impact, and availability.
Users are questioning whether the "Projects" feature in the ChatGPT web app and desktop app utilize different image generation models. Evidence suggests the web version produces higher-quality images and better understands prompts, such as generating multiple separate images without combining them into a collage. In contrast, the desktop app struggles with the same instructions, often failing to follow specific directives like avoiding collages.
With the exact same prompt, the web version understands my instructions very well and generates high-quality images. However, the desktop app seems to have much more difficulty understanding the same instructions.
For example, when I ask it to generate multiple separate images individually and explicitly tell it not to combine them into a collage, the web version follows the instruction perfectly. But the desktop app keeps generating the images as a collage.