Head to head: Wan vs OpenAI: GPT-5.4 Image 2
Wan vs OpenAI: GPT-5.4 Image 2
By Ryan Merket · Published
A test of image-making priorities: exact prompt fidelity and graphic restraint on one side, richer rendering and polished compositions on the other. The results turn on which kind of accuracy matters most across a varied set of briefs.
OpenAI: GPT-5.4 Image 2 takes the matchup, 61.9 to Wan’s 56.5, with five task wins to three. That’s a meaningful edge, but the limited confidence rating makes this a **lean**, not a runaway: the models trade strengths, and neither dominates every kind of brief. OpenAI’s clearest advantage is in scenes where convincing materials and spatial relationships matter. It wins the bedroom layout with the desk correctly placed under the window, and its kettle-and-tram scene makes the smoked glass, reflections, and puddle feel more convincing. It also comes out ahead on the reflective still life and multi-line text, though the text task is close: both models render the wording accurately. Wan is the more dependable choice when the prompt specifies a visual system. It wins the precise counter layout with cleaner isometric-vector styling, and its desk-kit and restricted-palette images better honor the requested flat look and wide format. OpenAI’s more photorealistic approach can be a liability here; square crops, gradients, or extra detail pull it away from the brief. The distinction is not simply polish versus quality: Wan is often more disciplined about style and aspect ratio, while OpenAI more often delivers the stronger overall scene. **Final call: OpenAI: GPT-5.4 Image 2 wins on balance, but only by a lean; choose Wan when strict format and graphic-style fidelity are the priority.**
Reflections & glass
A photorealistic close-up of a chrome teapot on a polished dark marble counter next to a glass of water; the window and a red apple on the counter must be correctly reflected in BOTH the chrome and the water's surface, dramatic side light, 16:9.
OpenAI: GPT-6 Luna: Model A better matches the requested 16:9 close-up and clearly includes the chrome teapot, water glass, dark marble, window, and apple, while Model B has more dramatic lighting and a stronger window reflection in the chrome. Neither clearly shows both the window and apple reflected on the water’s surface, and Model B also misses the requested aspect ratio. (Second judge pass, order swapped — scores are the average of both: Model A better matches the requested 16:9 composition and presents a cohe | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 frame and c…
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 B places the desk directly under the window and the floor lamp in the front-right corner, while Model A’s desk is offset from the window. Model A better matches the flat-vector style, but Model B has a more complete, polished composition with the requested spatial arrangement. (Second judge pass, order swapped — scores are the average of both: Model B closely matches the requested layout, with the bed against the left wall, a centered round rug, a desk directly beneath the back-wall window | Anthropic: Claude Fable 5.1: Model B places every element as specified—bed o…
precise-counter-layout
An isometric flat-lay illustration of a compact bathroom counter in crisp modern vector style: a mint electric toothbrush stands in a cup to the left of a round peach soap dish; a silver toothpaste tube lies horizontally behind the soap dish with its cap pointing right; a folded sand-colored washcloth is directly under the soap dish; a tiny cobalt mouthwash bottle is between the toothbrush cup and the soap dish; a white razor rests on top of the washcloth and to the right of the soap dish; and a black hair clip is in front of the soap dish but left of the razor; bright even lighting, immaculate geometry, exact spatial relationships.
OpenAI: GPT-6 Luna: Model A follows the requested crisp isometric vector style and places the specified items in the requested relationships. Model B has polished detail and a similar layout, but its photorealistic presentation and added soap bar make it less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model B gets the requested spatial relationships especially right, including the horizontal toothpaste tube behind the dish and the razor and clip positioned around the | Anthropic: Claude Fable 5.1: Model A matches the requested isometric vector…
mirror-kettle-window
A photorealistic 16:9 product-ad scene set in a narrow dawn kitchen: a smoked-glass electric kettle with a visible water line sits on a pale terrazzo counter beside a polished chrome manual coffee grinder and a small cobalt detergent bottle, with a rain-streaked window behind them showing a blurry orange tram outside; the kettle’s curved glass body must correctly refract the grinder and bottle through the water and glass, the chrome grinder must reflect the kettle, window light, and tram color as warped highlights, and a shallow puddle on the counter must mirror the lower edges of all three objects accurately; soft blue morning light, crisp focus on the products, realistic reflections and transparency.
OpenAI: GPT-6 Luna: Model B better captures the smoked-glass kettle, convincing orange-tram refraction, chrome reflections, and mirrored puddle, with stronger overall realism. Model A has the requested wide framing and pale terrazzo, but its glass and reflections are less convincing; Model B’s square framing and darker counter are notable deviations. (Second judge pass, order swapped — scores are the average of both: Model A better matches the requested 16:9 product-ad framing and presents the products cleanly agai | Anthropic: Claude Fable 5.1: Model B nails the smoked-glass kettle, blue daw…
color-bound-desk-kit
A clean studio product lineup in vibrant flat-vector design, 16:9: five everyday desk items arranged in a gentle arc on an off-white background, each with its own exact color-material pairing and no mixing: a matte teal aluminum water bottle, a glossy coral ceramic mug, a transparent amber acrylic ruler, a brushed violet stainless stapler, and a charcoal felt notebook with a lemon-yellow elastic band; add subtle drop shadows, a centered composition, and bright editorial lighting, making sure every named color stays attached only to its correct object and material.
OpenAI: GPT-6 Luna: Model A better matches the requested vibrant flat-vector style, 16:9 format, and gentle-arc lineup while keeping all five objects and their colors distinct. Model B renders the materials convincingly, but its photorealistic look, square framing, and large empty area above the objects depart from the prompt. (Second judge pass, order swapped — scores are the average of both: Model A better matches the requested 16:9 vibrant flat-vector editorial style while keeping all five items and their colors | Anthropic: Claude Fable 5.1: Model A honors the 16:9 ratio and flat-vector s…
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 required 16:9 format and flat, tightly restricted palette, though its campsite is cluttered and the floating circles are distracting. Model B is more polished and balanced, but its square format and gradient-like sky conflict with the prompt. (Second judge pass, order swapped — scores are the average of both: Model B is more visually cohesive, but its square format and visible background gradients conflict with the prompt. Model A has the requested 16:9 format and a fl | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 frame and s…
Legible multi-line text
A minimalist event poster with three lines of crisp, correctly-spelled text stacked and centered: 'NIGHT MARKET' large on top, 'Fridays · 6–11pm' in the middle, 'Riverside Pier 4' at the bottom, on a deep navy background, clean sans-serif, subtle grain.
OpenAI: GPT-6 Luna: Both models render all three lines accurately and center them on a deep navy background. Model A better matches the crisp, clean sans-serif and subtle-grain request, while Model B’s heavier grain visibly roughens the lettering. (Second judge pass, order swapped — scores are the average of both: Model B has crisp, correctly rendered text in a balanced, centered poster layout, with a restrained navy-and-cream palette and subtle grain. Model A also renders the wording accurately, but its wide forma | Anthropic: Claude Fable 5.1: Both Model A and Model B render all three lines…
cassette-player-palette
A retro-futurist poster-style illustration, 16:9, of a portable cassette player and bundled earbuds on a tabletop, restricted strictly to the named palette 'Harbor Dusk': navy, slate blue, seafoam, muted coral, warm sand, and ivory only, with no other hues anywhere in the image; use bold graphic shapes, soft grain, and a dramatic diagonal composition lit like late evening window light, ensuring every detail—including shadows, highlights, cord, buttons, tabletop, and background accents—stays entirely within that palette.
OpenAI: GPT-6 Luna: Model A follows the requested 16:9 format and presents the player, bundled earbuds, diagonal tabletop, evening window light, and restricted palette with little distraction. Model B has a more elaborate poster aesthetic, but its square format departs substantially from the prompt and its extra promotional copy and city scene make the composition less focused. (Second judge pass, order swapped — scores are the average of both: Model B has richer poster detail and dramatic lighting, but its square | Anthropic: Claude Fable 5.1: Model A honors the explicit 16:9 ratio and keeps…
Matchup powered by OpenRouter.