Which company has the best Image Edit AI end of September?

What this question means

Ninety percent for OpenAI, 8% for Google, and everyone else — Reve, Luma AI, Alibaba, Pruna, StepFun, SpaceXAI — at 1% or below across a fourteen-name field.

That is the most concentrated of the AI benchmark questions on this board, and the contrast with its siblings is the useful part. On general model quality the leader sits at 87%; on the Code Arena WebDev evaluation, a different company leads at 68%. Here a third configuration entirely, with an even larger margin.

Image editing is a narrower capability than general reasoning, and narrow capabilities tend to produce more decisive leaderboards. There is less room for a model to be better in some respects and worse in others, so the crowd can be confident in a way broader questions do not allow.

As with all of these, the resolution source decides the answer. "Best" means whatever the named benchmark says it means, and that choice matters more than any general impression of the field.