Head to head: Wan vs Google: Nano Banana Pro (Gemini 3 Pro Image)
Wan vs Google: Nano Banana Pro (Gemini 3 Pro Image)
By Ryan Merket · Published
Wan and Google’s Nano Banana Pro (Gemini 3 Pro Image) face off across eight image tasks, from exact object counts and poster text to action scenes and reflective surfaces. The comparison turns on a mix of reliable instruction-following, composition, and small but consequential detail errors.
Google’s Nano Banana Pro is the saved winner on this task sample, with aggregate scores of **64.2** to Wan’s **59.8**. The editorial assessment is a clear result, not a statistical claim: Google won four tasks, Wan won one, and three were ties. Google’s advantage came from several different kinds of prompt-following. It got the seven-cup count right, made the canal rescue feel more coherent and urgent, and showed the kayak spray skirt and gear bag more convincingly. It also edged Wan on the hands-and-anatomy task, where the warmer, more natural candid composition outweighed Wan’s cleaner hand rendering. Wan’s strongest result was the reflections-and-glass scene: it clearly placed the red apple on the counter alongside the teapot and glass, while Google showed the apple only as a reflection. The remaining comparisons were close enough to call ties: both models handled the three-line poster text, while the velodrome and boxer portrait traded strengths in perspective, framing, and detail. **Verdict: Google: Nano Banana Pro (Gemini 3 Pro Image) wins this matchup, but the lead rests on a varied set of task-level advantages rather than a sweep. This assessment applies to the saved task sample.**
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 presents exactly seven clearly separated, highly uniform espresso cups in a straight row on white marble, with crisp overhead lighting. Model B also meets the count and arrangement well, but the espresso surfaces and cup appearance are slightly less consistent. (Second judge pass, order swapped — scores are the average of both: Model B presents exactly seven clearly separated, evenly spaced cups in a clean overhead row, with soft lighting and restrained marble veining. Model A also gets | Anthropic: Claude Fable 5.1: Both Model A and Model B deliver exactly seven i…
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: Model A is clean and well balanced, with all three lines rendered correctly. Model B gives the headline stronger emphasis while keeping the centered text crisp and legible, making for a more striking poster overall. (Second judge pass, order swapped — scores are the average of both: Both render all three lines clearly and correctly on a deep navy, subtly grained background. Model A has a more balanced, restrained minimalist layout, while Model B’s oversized title and supporting text feel somewha | Anthropic: Claude Fable 5.1: Model A renders all three lines correctly with…
Canal rescue scramble
Documentary photojournalism image of an urban rescue-kayak race accident beneath the Kestrel Lock footbridge, 16:9: five subjects interacting in one coherent scene—a drenched paddler hauling a lime kayak onto the concrete edge, a teammate kneeling to clip a tow line, a race marshal in an orange drysuit reaching down from the dock, a paramedic opening a silver trauma bag, and a soaked border collie shaking water nearby—balanced composition with the bridge framing the action, rippling canal reflections, scattered paddles and numbered bibs, overcast light with one bright shaft of sun breaking through.
OpenAI: GPT-6 Luna: Model B captures the rescue as a more dynamic, coherent scene, with the marshal reaching down, the paramedic opening the trauma bag, and the soaked dog visibly shaking beneath the bridge. Model A includes most requested elements, but the action feels less urgent and the bridge sign is misspelled. (Second judge pass, order swapped — scores are the average of both: Model B more clearly depicts the requested rescue roles interacting, with the marshal reaching down, the paramedic opening a trauma ba | Anthropic: Claude Fable 5.1: Model B nails nearly every element: correct 'KE…
Velodrome sprint perspective
A documentary-style wide shot inside the fictional Brno East Velodrome during a match sprint, 16:9: the camera is low on the inner apron looking down the steeply banked track so the lane lines and railings converge to a single believable vanishing point; in the foreground a cyclist in a silver skinsuit and teal helmet is large and sharply detailed, midground two riders in amber and plum kits are clearly smaller and aligned to the curve, and in the distant upper bend a tiny official stands near the rail; correct relative scale, no warped bicycles, no impossible banking, late-afternoon skylight spilling through roof panels.
OpenAI: GPT-6 Luna: Model A better captures the low, wide velodrome perspective, with the requested rider scale hierarchy, amber and plum competitors, and a tiny official near the rail. Model B has strong documentary lighting, but its foreground apron and less clearly converging track lines weaken the intended composition. (Second judge pass, order swapped — scores are the average of both: Model B better matches the requested foreground-to-background scale, with the silver-and-teal rider prominent, the two racers s | Anthropic: Claude Fable 5.1: Model B nails the low apron viewpoint, convinci…
Boxer hand wrap portrait
A cinematic ringside portrait of a female southpaw boxer in the Blue Furnace Gym, shown from mid-thigh up under hard overhead training lights: she is tightening the red wrap on her left hand with her right hand, fingers clearly visible and anatomically correct, natural shoulders and torso proportions, a light sheen of sweat, taped heavy bags and a hand-painted round clock blurred in the background, realistic skin texture, shallow depth of field, gritty sports photography style.
OpenAI: GPT-6 Luna: Model A better matches the requested mid-thigh-up portrait, keeps the boxer’s full head in frame, and clearly shows her tightening the red wrap with visible fingers. Model B has a grittier cinematic look and more explicit Blue Furnace branding, but the cropped head and less balanced framing weaken the portrait. (Second judge pass, order swapped — scores are the average of both: Model A gives a clearer, more complete portrait with the wrapping action and both hands easy to read, while retaining t | Anthropic: Claude Fable 5.1: Model A delivers a true ringside portrait with…
Hands & anatomy
A natural candid photo of a person tying a friendship bracelet, both hands clearly visible with anatomically correct fingers and a plausible grip on the threads, soft daylight, shallow depth of field, 16:9.
OpenAI: GPT-6 Luna: Model A has especially clear, anatomically convincing hands and a plausible grip, while Model B more fully captures the candid, cheerful moment and shows the friendship bracelet in context. Model B’s hands are slightly less cleanly rendered, but its composition and overall warmth give it the edge. (Second judge pass, order swapped — scores are the average of both: Model B feels more like a natural candid moment, with both hands visibly working the colorful threads and a pleasing soft-focus garde | Anthropic: Claude Fable 5.1: Model A delivers a crisp close-up with clearly…
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 clearly shows the red apple on the counter alongside the teapot and glass, with convincing window reflections in the chrome; Model B has dramatic lighting and polished reflections, but the apple appears only as a reflection rather than clearly sitting on the counter. Neither image clearly shows the apple reflected in the water’s surface as requested. (Second judge pass, order swapped — scores are the average of both: Model A clearly shows the teapot, glass, and red apple together, with s | Anthropic: Claude Fable 5.1: Model A shows the apple on the counter but the…
Kayak gear binding
A crisp studio product photograph of a sprint-kayak pit setup on a matte charcoal floor, shot slightly from above in 16:9: a cobalt-blue carbon-fiber racing kayak resting beside a saffron-yellow neoprene spray skirt, a jade-green anodized aluminum paddle with black blades, a cherry-red glossy fiberglass helmet, a white mesh gear bag with a brass zipper, and a violet silicone water bottle; every item must keep its own exact color and material with no color bleeding or swaps, ultra-detailed commercial lighting with softbox reflections and clean shadows.
OpenAI: GPT-6 Luna: Model B more clearly depicts the spray skirt and mesh gear bag, and its charcoal-floor arrangement has a polished commercial studio look. Model A includes the requested color lineup, but the skirt and bag read less accurately and the floor is noticeably lighter. (Second judge pass, order swapped — scores are the average of both: Model B more closely matches the requested gear and materials, especially the clearly recognizable spray skirt, while presenting all items in a polished, balanced studio | Anthropic: Claude Fable 5.1: Model B renders every item correctly — a true s…
Matchup powered by OpenRouter.