Head to head: FLUX 2 vs Google: Nano Banana Pro (Gemini 3 Pro Image)
FLUX 2 vs Google: Nano Banana Pro (Gemini 3 Pro Image)
By Ryan Merket · Published
A prompt-following face-off across typography, spatial layout, constrained styles, object relationships, and exact counting. The comparison weighs clean instruction execution against richer image detail across eight tasks.
FLUX 2’s strongest results are practical: it nails the requested three-line poster layout and shows exactly seven espresso-filled cups. Google: Nano Banana Pro (Gemini 3 Pro Image) takes four tasks, including perspective, restricted palette, named art style, and attribute binding, where judges favored its clearer execution or richer rendering. Spatial layout and negation are ties, with shortcomings noted in both images on the no-plants request. The saved aggregate scores are **59.8 for FLUX 2** and **64.8 for Google: Nano Banana Pro (Gemini 3 Pro Image)**. That result supports a **slight edge** for Google in this matchup—not a decisive separation: FLUX wins two tasks, Google wins four, and two are ties. The tradeoff is specific: FLUX is more dependable on exact text layout and countable, correctly filled objects, while Google’s wins cluster around visual style, scene depth, and polished material detail. This assessment applies to the saved task sample, not image-model performance in general. **Final call: Google: Nano Banana Pro (Gemini 3 Pro Image), by a slight edge on this sample.**
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 presents the exact three requested lines, centered and legible on a deep navy, subtly textured background. Model B renders the wording correctly but splits the headline across two lines and places the poster on a light mockup background with an awkwardly large gap. (Second judge pass, order swapped — scores are the average of both: Model A follows the requested three-line centered layout with all text crisp and correctly spelled on a deep navy background. Model B renders the wording accu | Anthropic: Claude Fable 5.1: Model A delivers exactly what was asked: three…
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 places the bed against the left wall, centers the round rug, and clearly positions the desk beneath the back-wall window and the lamp in the front-right corner. Model B is attractive and follows most of the layout, but its rug sits somewhat forward and its arched lamp is less clearly contained in the corner. (Second judge pass, order swapped — scores are the average of both: Model B places the bed, centered rug, desk beneath the back-wall window, and floor lamp in the front-right area mo | Anthropic: Claude Fable 5.1: Model A places the bed cleanly against the left…
Negation
A cozy reading nook with an armchair, a stack of books, and a mug of tea by a window — with absolutely NO plants, NO lamps, and NO artwork or picture frames anywhere in the frame. Warm afternoon light, 16:9.
OpenAI: GPT-6 Luna: Model A better honors the strict no-plants constraint, though a leafy shape is visible at the lower-left edge; Model B shows abundant garden vegetation through the window. Both have the requested cozy chair, books, tea, and warm light, while Model B offers a more expansive room composition. (Second judge pass, order swapped — scores are the average of both: Model A presents the requested chair, books, tea, and warm window light without any visible lamps or artwork, and its view outside is less c | Anthropic: Claude Fable 5.1: Model A violates the negation by including a vi…
Perspective & scale
A photorealistic one-point-perspective shot down a long empty library aisle, shelves receding to a single vanishing point, books getting consistently smaller with distance and nothing warped or wrongly sized, warm overhead light, 16:9.
OpenAI: GPT-6 Luna: Model A more clearly delivers a single, centered vanishing point with a long empty aisle and consistently diminishing shelves and books. Model B has richer photorealistic detail, but its foreground shelving makes the aisle feel less cleanly defined and the perspective less direct. (Second judge pass, order swapped — scores are the average of both: Model B combines a convincing single vanishing point and clear scale recession with richer, more photorealistic library detail and warm overhead light | Anthropic: Claude Fable 5.1: Model A delivers a textbook one-point perspecti…
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 B adheres more closely to the four-color, flat-vector constraint, while Model A uses gradients and many tonal variations despite its appealing campsite scene. Model A has a more natural, balanced composition; Model B is cleanly rendered but its heavy horizontal bands make the scene feel unusually divided. (Second judge pass, order swapped — scores are the average of both: Model B more closely follows the strict four-color, flat-vector requirement, while Model A uses prominent gradients and | Anthropic: Claude Fable 5.1: Model A is a polished, well-balanced campsite s…
Named art style
A ukiyo-e woodblock print of a fishing boat riding a large cresting wave at dawn, faithful to the flat color planes, bold outlines, and stylized foam of the tradition, muted indigo and cream palette.
OpenAI: GPT-6 Luna: Model B more clearly captures the ukiyo-e tradition with bold outlines, stylized foam, a muted indigo-and-cream palette, and a dawn sun, while its boat rides directly within the cresting wave. Model A has a striking wave and fitting palette, but the boat feels less integrated into the action and the composition is more crowded. (Second judge pass, order swapped — scores are the average of both: Model B more directly depicts a fishing boat riding the cresting wave, with the dawn, muted indigo-and | Anthropic: Claude Fable 5.1: Model B delivers a convincingly authentic ukiyo…
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: Both images correctly place the yellow duck atop the green cube, with the red sphere to its right and the blue cylinder behind. Model B is more convincingly photorealistic, with richer material detail and a stronger studio-lit composition. (Second judge pass, order swapped — scores are the average of both: Model B more convincingly delivers the photorealistic tabletop scene and soft studio lighting, with all four objects and their color and placement relationships clear. Model A follows the arra | Anthropic: Claude Fable 5.1: Both bind colors correctly and place the green…
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 separated, identical espresso cups in a straight row on white marble, with espresso visible in each cup. Model B also has seven cups in a clean row, but the cups are empty and the image is more washed out. (Second judge pass, order swapped — scores are the average of both: Model B and Model A both show exactly seven individually countable cups in a straight overhead row on white marble. Model A’s espresso-filled cups are more recognizable and visually distinct | Anthropic: Claude Fable 5.1: Both Model A and Model B deliver exactly seven…
Matchup powered by OpenRouter.