>>106283486
> I half suspect that they might be running it on very high parameters/settings for this blind testing phase.
Could be, but it's miles better than Imagen 4 and gpt1-image