Kitchen dialogue
Focus falls off towards the camera as well as away from it. A model that reads blur as distance has to put the foreground head somewhere behind the window.
Table of Contents
The picture

Where it shows
The difficulty is not spread evenly over the frame. Each link opens the comparison already zoomed to one of these places.
How the 13 shipped models read it
Each at its finest setting. The differences are the point - and the reason there is no score under any of them: a generated image has no ground-truth depth to score against, so what a map gets right is a judgment, not a number.
Depth Anything V2 Small518×518
Depth Anything V2 Small518×518
Depth Anything V2 Base518×518
Depth Anything V2 Base518×518
Depth Anything V2 Large518×518
Depth Anything V3 Mono Large1036×1036
Depth Anything V3 Mono Large1036×1036
Lotus-D1024×1024
Depth Anything V1 Small518×518
Depth Anything V1 Base518×518
MiDaS v2.1 Small256×256
MiDaS DPT-Swin2-Tiny256×256
MiDaS DPT-Hybrid384×38463 further models are on this scene too - they are in the comparison rather than here, because a page of 76 pictures is a page nobody reads.
The other 24 scenes
Alpine valleyAnime, cherry blossomsAnime, ramen barBlizzard whiteoutCG animation, bedroomClaymation kitchenDancer in confetti1950s sitcomFisherman portraitHorror forestMacro, beeMetro corridorMotorcycle chaseNews studioOverhead cookingParking garage lineupPixel art castlePrison mesh gatesPure blackPure whiteRainy night streetWatercolor illustrationWedding, soap bubblesWestern duel