Head to head: Wan vs OpenAI: GPT-5.4 Image 2
Wan vs OpenAI: GPT-5.4 Image 2
By Ryan Merket · Published
Wan and OpenAI: GPT-5.4 Image 2 split this image-model sample between tight prompt constraints and more elaborate scene-making. The comparison spans layout, counting, object relationships, text, and visual storytelling.
OpenAI: GPT-5.4 Image 2 is the saved winner, with an aggregate score of **63.4** to Wan’s **56.5** and five task wins to three. The gap is an editorial **slight edge (not statistical)**, not a sweep: Wan won exact counting, attribute binding, and restricted palette, where correct quantities, spatial relationships, aspect ratio, and limited colors mattered most. OpenAI won the rainy ranger workbench, relieved snow leopard, estuary rescue poster, and precise marsh still life, as well as spatial layout. Its strongest results were more compelling scene details and, on the poster, cleanly rendered required wording without extra copy. But square crops repeatedly weakened otherwise polished images, especially when the prompt specified 16:9. Wan’s results were more dependable on several literal constraints: it kept the seven-cup image wide, preserved the tabletop colors and placements, and better respected the campsite’s flat, restricted palette. OpenAI’s stronger storytelling did not erase those misses; Wan’s constraint discipline did not make it the more frequent task winner. **Final call: OpenAI: GPT-5.4 Image 2 takes this matchup by a slight edge, with five wins to Wan’s three. This assessment applies only to the saved task sample.**
Spatial layout
A clean isometric illustration of a bedroom: a bed against the LEFT wall, a round rug centered on the floor, a desk under the WINDOW on the back wall, and a floor lamp in the FRONT-RIGHT corner. Flat-vector style, consistent perspective.
OpenAI: GPT-6 Luna: Model A follows the requested flat-vector style and places the bed, centered round rug, desk beneath the back-wall window, and lamp in the front-right corner. Model B matches the layout and is highly polished, but its more dimensional, shaded rendering is less faithful to the specified flat-vector look. (Second judge pass, order swapped — scores are the average of both: Model B places the bed, centered round rug, desk directly under the back-wall window, and floor lamp in the front-right area, w | Anthropic: Claude Fable 5.1: Both place the bed on the left wall, the rug ce…
Exact counting
A clean overhead flat lay on white marble of EXACTLY seven identical espresso cups arranged in a single straight row, each clearly separate and individually countable, even soft lighting, 16:9.
OpenAI: GPT-6 Luna: Model A shows exactly seven clearly countable cups in a straight row and matches the requested 16:9 framing. Model B is also polished and accurately counted, but its square format misses the specified aspect ratio. (Second judge pass, order swapped — scores are the average of both: Model B shows exactly seven distinct cups in a clean, evenly spaced row, but its square format misses the requested 16:9 aspect ratio. Model A also has exactly seven countable cups and matches the requested wide forma | Anthropic: Claude Fable 5.1: Model A delivers exactly seven identical espres…
Rainy ranger workbench
Photorealistic 16:9 documentary-style scene inside a remote wetland field station at blue-hour dusk: a wildlife ranger leans over a scarred oak workbench repairing a silver aluminum radio collar beside a steaming enamel mug, a folded moss-green canvas jacket, and a damp leather field notebook, while rain-speckled windows reveal blurred reeds outside; mixed lighting from a warm tungsten desk lamp and cool cyan dusk creates believable reflections on brushed metal screws, soft sheen on wet wood, natural skin texture on the ranger’s hands and face, and realistic fabric weave on the jacket, with physically accurate shadows, subtle ambient bounce light, and shallow depth of field focused on the collar.
OpenAI: GPT-6 Luna: Model B more clearly depicts the ranger repairing a distinct silver radio collar, with steam, rain-speckled windows, and focused mixed lighting; Model A better matches the requested 16:9 framing but its device reads more like a small radio than a collar. Model B’s square crop and awkward notebook lettering are drawbacks, but its subject and scene are more compelling overall. (Second judge pass, order swapped — scores are the average of both: Model B closely matches the requested scene, with the | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 frame with r…
Relieved snow leopard
Ultra-detailed photorealistic close-up portrait of a snow leopard at first sunrise on a rocky Himalayan ledge, captured with a long lens in 16:9, the animal convincingly expressing unmistakable relief after finding its cub: softened eyes with slight moisture, relaxed ears, mouth gently parted as it nuzzles the top of a small cub’s head, facial muscles loose rather than tense, body posture releasing tension; golden rim light catches individual whiskers and thick fur, cold blue mountain shadows behind, cinematic but natural wildlife photography.
OpenAI: GPT-6 Luna: Model A better matches the requested 16:9 close-up and shows the gently parted mouth, golden rim light, and nuzzling interaction. Model B conveys a softer, more clearly relieved bond and has a pleasing composition, but is less faithful to the specified framing and mouth expression. (Second judge pass, order swapped — scores are the average of both: Model A better matches the requested 16:9 close-up and clearly shows the mother nuzzling her cub with her mouth gently parted. Model B has a beautifu | Anthropic: Claude Fable 5.1: Model A honors the 16:9 close-up framing and go…
Attribute binding
A photorealistic tabletop scene: a matte green cube to the LEFT of a glossy red sphere, a small blue cylinder BEHIND both, and a single yellow rubber duck sitting ON TOP of the cube. Colors must stay on the correct objects. Soft studio light, 16:9.
OpenAI: GPT-6 Luna: Model A follows the requested 16:9 tabletop composition and clearly preserves the object colors and spatial relationships. Model B is polished and similarly accurate, but its square framing misses the specified aspect ratio. (Second judge pass, order swapped — scores are the average of both: Model A follows the object placements and colors while also matching the requested 16:9 tabletop composition; Model B is polished but uses a square crop and a more reflective-looking tabletop. Both clearly s | Anthropic: Claude Fable 5.1: Model A satisfies every constraint: green matte…
Estuary rescue poster
Clean graphic design poster in a realistic photographed bus-stop lightbox, 16:9, promoting a coastal wildlife rescue event with a crisp illustrated osprey diving over dark teal water; the typography must be perfectly legible and exactly read: “BRACKEN TIDE BIRD CLINIC” on the top line, “OPEN SAT 14 MAY” beneath it, and “LOWER MERE QUAY” at the bottom; use bold condensed sans-serif lettering in off-white with a small coral accent bar, balanced modern layout, straight-on composition, even illumination, and no extra words anywhere.
OpenAI: GPT-6 Luna: Model B renders the three requested lines cleanly and avoids extra poster copy, with a striking, crisp osprey illustration; its main shortcoming is the square format rather than the requested 16:9. Model A matches the requested wide format and has a balanced layout, but adds extra text to the poster. (Second judge pass, order swapped — scores are the average of both: Model B renders all required wording exactly, keeps the requested hierarchy, and pairs it with a striking illustrated osprey in a | Anthropic: Claude Fable 5.1: Model A nails the 16:9 lightbox format and clean…
Restricted palette
A flat-vector illustration of a mountain campsite at dusk using ONLY four colors — cream, burnt orange, teal, and charcoal — with no gradients and no stray off-palette colors, balanced composition, 16:9.
OpenAI: GPT-6 Luna: Model A better follows the strict four-color, flat-vector requirement and the requested 16:9 format, while clearly depicting a campsite. Model B is more polished and dramatic, but it is square and uses gradients and apparent off-palette colors. (Second judge pass, order swapped — scores are the average of both: Model A follows the requested 16:9 format and uses a mostly flat, limited palette, while Model B has a more dramatic scene but is square and visibly uses gradients. Model B is more visual | Anthropic: Claude Fable 5.1: Model A delivers the required 16:9 frame and st…
Precise marsh still life
Highly realistic 16:9 nature still life on a muddy marsh bank under soft overcast light: a red canoe paddle lies horizontally, a glass jar of fireflies sits on top of the paddle near its right end, a pair of yellow binoculars is to the left of the jar, a curled fern frond is behind the jar, a smooth black river stone is between the binoculars and the jar, and a blue kingfisher feather rests in front of the paddle; the layout must clearly show every spatial relationship exactly as stated, viewed from a slightly elevated three-quarter angle with crisp depth and natural textures.
OpenAI: GPT-6 Luna: Model A more closely matches the requested 16:9 framing and presents all objects clearly in the specified spatial relationships. Model B depicts the curled fern and fireflies more convincingly, but its square framing departs from the prompt and its darker composition is less legible. (Second judge pass, order swapped — scores are the average of both: Model A follows the requested 16:9 format and makes the object layout easy to read, though its fern is not curled and the fireflies look like brigh | Anthropic: Claude Fable 5.1: Model A honors the 16:9 format and marsh settin…
Matchup powered by OpenRouter.