Skip to main content

Reality checks

Can AI actually do this?

Each case is one real image and a question with a fixed answer key, written before any model answered. You can check every verdict with your own eyes.

Solved

Every tested model, top and budget, got every check right.

Top models only

The top models got it right; at least one budget model did not.

Hit & miss

Results were mixed, even among top models.

Not yet

No tested model got it right.

What the categories test