ImageBench

RealBench V1

A visual Turing test for image models: can their output pass as a real photograph? Each model's realism score is the share of human votes that judged its AI images as real — voted on by the ImageBench community in a quick real-vs-AI game.

Realism leaderboard

Sorted most photoreal first. Models need at least 10 votes to appear. Today's most convincing model is fal/google/nano-banana-pro.

#ModelRealism scoreRated real / votesImages
1fal/google/nano-banana-pro
41%
442 / 1,08185
2local/z-image-turbo-6b
35%
274 / 77961
3fal/google/nano-banana-2
35%
381 / 1,08585
4bfl/flux-2-max
33%
795 / 2,388189
5bfl/flux-2-pro
27%
287 / 1,06985
6fal/bfl/flux-1-dev
25%
204 / 82865
7fal/fast-sdxl
25%
16 / 655
8openai/gpt-image-2
23%
248 / 1,08685
9bfl/flux-2-flex
20%
168 / 82265
10local/sana-1.5-1.6b
7%
56 / 82565

10,028 votes · snapshot from 6/15/2026