gpt-imageIMAGE

Qwen-Image 2.1 vs GPT-Image-2.5 on the same prompt

A same-prompt comparison of locally deployed Qwen-Image 2.1 and GPT-Image-2.5 finds GPT richer in detail but fragmented, Qwen overly smudged, and the local model impressively capable.

@akokoi1Archived 2026.09.23Originally posted 2026.09.22

Qwen-Image 2.1 vs GPT-Image-2.5 on the same prompt (image 1 of 2)

Qwen-Image 2.1 vs GPT-Image-2.5 on the same prompt (image 2 of 2)

WY @akokoi1Originally posted 2026.09.22

Open original on X

同样的提示词，左边是本地部署的Qwen-image-2.1，右边是GPT-Image-2.5。 GPT的细节很丰富，但是画面依旧有点碎，Qwen的涂抹感又有点重。说实话作为消费级显卡就能部署的本地模型，很能打了。期待Qwen-image下一个开源版本。提示词放评论区⬇️

This post quotes another post: x.com/akokoi1/status/2101865226547478950

x.com/akokoi1/status/2102319136307363873Archived 2026.09.23 10:15