Gemini 4 Pro trails 3.7 Flash on visual reasoning, Burkov says

Burkov says the model is not the strongest in the world after all and, in his tests, is clearly weaker than Gemini 3.7 Flash and Luna on visual reasoning tasks.

BURKOV

@burkovArchived 2026.09.22Originally posted 2026.09.22

BURKOV @burkovOriginally posted 2026.09.22

Open original on X

Apparently not the strongest model in the world after all :-) Also, based on my tests, sucks in visual reasoning tasks compared to Gemini 3.7 Flash and Luna.

This post quotes another post: x.com/spacexai/status/2102069815225586149

x.com/burkov/status/2102163647338360931Archived 2026.09.22 10:19

Open original on X