I've re-run my benchmark with the Flux 2 Pro model and found that in some cases the higher resolution models (I believe Flux 2 Pro handles 4k) can actually backfire on some of the tests because it'll introduce the equivalent of an almost ESRGAN style upscale which may add in unwanted additional details. (See the Constanza test in particular).

https://genai-showdown.specr.net/image-editing

▲

minimaxir 5 hours ago | parent [-]

That Constanza test result is baffling.

	▲	vunderba 4 hours ago \| parent [-]
		Agreed - I was quite surprised. Even though its a bog-standard 1024x1024 image, the somewhat low quality nature of a TV still provides for an interesting challenge. All the BFL models (Kontext Max and Flux 2 Pro) seemed to struggle hard with it.