Head to head: Wan vs Google: Nano Banana 2 (Gemini 3.1 Flash Image)
Wan vs Google: Nano Banana 2 (Gemini 3.1 Flash Image)
By Ryan Merket · Published
This matchup tests whether exact instruction following can outweigh broader strength in composition, spatial reasoning, object binding, and atmosphere. The eight tasks expose a clear difference between a specialist with notable peaks and a more dependable all-rounder.
Google’s Nano Banana 2 takes this matchup decisively: **65.0 to 57.2**, with six task wins to Wan’s two. The statistical verdict gives Google a **91% probability of winning**, making this a clear result rather than a narrow points victory. Google built its lead through range. It won the spatial-layout test by correctly positioning the desk beneath the window and the lamp in the front-right corner, handled attribute binding with better size fidelity, and rendered all six harbor objects as distinct items—including the standalone ivory oar lock Wan muddled into a larger paddle assembly. It also won both atmospheric poster tasks, producing the stronger rain-cooled Lake Iseo scene and the more convincing five-color Sirimiri travel print. Wan’s two wins were meaningful and reveal where it remains dangerous. In the desert-camp layout, it respected the unusually exact relationships—most importantly, putting the telescope on the crate and the kettle between the crate and mat—while Google introduced a ground-standing tripod and extra clutter. Wan was also plainly better at the multiline poster, delivering exactly three centered, correctly spelled lines where Google split “NIGHT MARKET” across two lines and weakened the requested presentation. Those flashes of precision do not offset Google’s broader command of layout, binding, materials, palette-driven illustration, and mood. **Final call: Google Nano Banana 2 wins clearly, 65.0–57.2, with a tentative lean.**
Precise Desert Camp Layout
In a sunlit 16:9 travel scene set on the red gravel plain of Wadi Rum, depict a small expedition camp with exact spatial relationships: a silver Land Rover is left of a striped beige tent; a blue water crate is directly in front of the tent; a brass telescope is on top of the crate; a rolled turquoise sleeping mat is right of the tent; a black kettle sits between the crate and the sleeping mat; and a rust-red sandstone arch rises behind the entire camp; realistic lighting, clear mid-distance composition, all objects fully visible and placed exactly as described.
OpenAI: GPT-5.6 Sol Pro: Model A follows the exact object relationships most closely, especially placing the brass telescope on top of the crate and the kettle between the crate and mat. Model B is highly attractive and realistic, but the telescope is supported by a separate ground-standing tripod rather than sitting on the crate, and extra camp items reduce the layout’s precision. (Second judge pass, order swapped — scores are the average of both: Model A more precisely places the brass telescope on top of the blue cra | Anthropic: Claude Fable 5.1: Model A places every object exactly as spe…
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-5.6 Sol Pro: Model A adheres much more closely to the restricted four-color, flat-vector requirement, though it appears to use slightly different teal tones and has a somewhat cluttered foreground. Model B has stronger composition and visual polish, but its multiple teal, orange, and dark shades clearly violate the strict palette constraint. (Second judge pass, order swapped — scores are the average of both: Model B delivers a polished, balanced dusk campsite with strong depth and visual appeal, though visib | Anthropic: Claude Fable 5.1: Model A holds tightly to the four-color re…
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-5.6 Sol Pro: Model B accurately places the bed along the left wall, the rug centrally, the desk directly under the back-wall window, and the lamp in the front-right corner, with polished detail and a balanced layout. Model A is clean and consistent, but its desk is beside rather than under the window and its rug sits somewhat forward of center. (Second judge pass, order swapped — scores are the average of both: Model B accurately places the bed against the left wall, the rug centrally, the desk directly bene | Anthropic: Claude Fable 5.1: Model A delivers a true isometric room wit…
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-5.6 Sol Pro: Model A binds every color and object correctly, but the blue cylinder appears somewhat large relative to the requested small size. Model B also satisfies all spatial and color constraints while depicting the cylinder more clearly as small, with especially natural lighting and depth. (Second judge pass, order swapped — scores are the average of both: Model B cleanly binds every color and spatial relationship while presenting the requested objects with especially convincing photorealism, soft ligh | Anthropic: Claude Fable 5.1: Both bind colors correctly and place the d…
Harbor Gear Binding
At dawn on a misty basalt harbor in the Faroe-like village of Norðvik, create a highly detailed claymation-style still life on a weathered pier: a matte teal ceramic lantern, a glossy saffron-yellow rubber rain boot, a brushed copper compass, a frosted lilac glass buoy, a charcoal wool cap, and an ivory-painted wooden oar lock, each object clearly separate and keeping its own exact color and material with no color bleeding or swapping; soft cold sunrise light, shallow sea haze, crisp textures, cinematic 16:9 composition.
OpenAI: GPT-5.6 Sol Pro: Model B more faithfully presents all six requested objects as clearly separate items, especially the ivory-painted wooden oar lock, while maintaining convincing colors, materials, misty harbor atmosphere, and a balanced cinematic arrangement. Model A is attractive and detailed, but substitutes a full paddle-and-hardware assembly for the standalone oar lock and has a somewhat more crowded foreground. (Second judge pass, order swapped — scores are the average of both: Model B depicts all six reque | Anthropic: Claude Fable 5.1: Model B renders all six objects accurately…
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-5.6 Sol Pro: Model A precisely delivers three crisp, correctly spelled, centered lines on a deep navy, subtly grained background with strong hierarchy. Model B renders the text correctly and looks polished, but wraps “NIGHT MARKET” onto two lines and presents the poster against a white surround, reducing prompt adherence. (Second judge pass, order swapped — scores are the average of both: Model A precisely delivers the requested three centered lines with crisp, correctly spelled text, clear hierarchy, deep n | Anthropic: Claude Fable 5.1: Model A delivers exactly three crisp, corr…
Melancholy Funicular Vector
A clean vector illustration of an old mountain funicular climbing through terraced cypress hills above Lake Iseo at blue hour, with a wistful, rain-cooled mood: simplified geometric forms, crisp linework, subtle grain, glowing carriage windows, a lone traveler holding a folded map on the platform, reflections on damp rails, minimal but expressive background, polished editorial style.
OpenAI: GPT-5.6 Sol Pro: Model B more fully realizes the Lake Iseo setting, terraced cypress hills, rain-cooled atmosphere, damp reflections, subtle grain, and lone traveler while offering a more layered editorial composition. Model A is clean and attractive with strong glowing windows, but its simplified setting and drier, less melancholic atmosphere capture fewer prompt-specific details. (Second judge pass, order swapped — scores are the average of both: Model B more fully captures Lake Iseo, terraced cypress hills, d | Anthropic: Claude Fable 5.1: Model B nails the wistful, rain-cooled blu…
Sirimiri Palette Only
A wide 16:9 travel poster illustration of the cliffside town of Lekeitio during a light sirimiri drizzle, rendered strictly and exclusively in the named Japanese traditional color palette 'Shinbashi-iro, Aonibi, Kokiake, Kinari, and Sumi'—no other hues anywhere in the image; clean graphic shapes, screen-printed texture, wet cobblestones, fishing boats below, low overcast light, restrained elegant composition.
OpenAI: GPT-5.6 Sol Pro: Model B more convincingly delivers a restrained screen-printed travel poster with visible drizzle, wet reflective surfaces, fishing boats, cliffside architecture, and a cohesive five-color appearance. Model A is attractive and clean but feels flatter, shows the drizzle less clearly, and includes distracting foreground objects that weaken the elegant composition. (Second judge pass, order swapped — scores are the average of both: Model B adheres more closely to the exclusive five-color palette wh | Anthropic: Claude Fable 5.1: Model B hews more tightly to the five-colo…
Matchup powered by OpenRouter.