Head to head: FLUX 2 vs Google: Nano Banana 2 (Gemini 3.1 Flash Image)
FLUX 2 vs Google: Nano Banana 2 (Gemini 3.1 Flash Image)
By Ryan Merket · Published
A side-by-side look at FLUX 2 and Google’s Nano Banana 2 across eight image-generation tasks, from precise object placement and readable text to atmospheric, material-rich scenes. The prompts test both visual polish and how faithfully each model handles a crowded brief.
The saved result favors Google’s Nano Banana 2: it scores **67.0** to FLUX 2’s **56.9**, with six task wins, no losses, and two ties. The editorial strength is **decisive** for this task sample—not a claim about statistical confidence or performance beyond it. Nano Banana’s advantage is mainly in following the brief while keeping scenes legible. It places the tabletop objects correctly, makes the bedroom layout easy to read, and better conveys grief in the orbital greenhouse scene. It also handles the chrome, glass, and reflected light in the still life more convincingly. The more demanding scenes make the difference especially clear. In the rain bazaar, Nano Banana depicts the requested brass-finger fitting, while FLUX reverses the interaction and shows a full prosthetic arm. In the neon alley, Nano Banana reads more clearly as a courier stopped on a scooter; FLUX brings strong atmosphere, but the action is less distinct. The two ties are useful counterpoints: FLUX more directly captures the crystal causeway, while both models render the requested three lines of text correctly. **Verdict: Google’s Nano Banana 2 is the stronger pick on this saved task sample, winning six tasks and tying two; FLUX’s best moments don’t offset its weaker prompt fidelity across the set.**
Carless Crystal Causeway
A wide establishing shot of the crystal causeway leading into the fantasy trade city of Halcyon Spire at dawn, lined with market pennants, porters, handcarts, tethered feathered pack-lizards, and mirrored obelisks catching pale peach light—but absolutely no cars, no wagons, and no motor vehicles of any kind anywhere in the scene; detailed epic fantasy matte-painting style, clear atmospheric depth, balanced panoramic composition, 16:9
OpenAI: GPT-6 Luna: Model A more directly captures the crystal causeway, peach dawn, pennants, porters, and distinctive feathered pack-lizards in a balanced panoramic composition. Model B is richly detailed and includes the requested trade bustle, but its crowded bridge and less clearly crystal-focused setting make it a slightly weaker match. (Second judge pass, order swapped — scores are the average of both: Model B more fully realizes the busy crystal causeway with porters, pennants, handcarts, feathered pack bea | Anthropic: Claude Fable 5.1: Model B delivers a true wide establishing shot…
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 B looks more photorealistic and clearly places the blue cylinder behind the cube and sphere, while preserving the requested object colors and duck-on-cube arrangement. Model A is cleanly composed, but its simpler, more CG-like rendering is less convincing as a photorealistic tabletop scene. (Second judge pass, order swapped — scores are the average of both: Model B clearly places the yellow duck atop the matte green cube, with the glossy red sphere to its right and the blue cylinder behind | Anthropic: Claude Fable 5.1: Model B satisfies every spatial relation convin…
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 key furnishings well, but Model B more clearly matches the requested flat-vector style while keeping the bed, centered rug, desk beneath the back-wall window, and lamp in the front-right area easy to read. Model B also has a more balanced, detailed composition. (Second judge pass, order swapped — scores are the average of both: Model B more closely matches the requested flat-vector style and clearly places the bed, centered rug, desk beneath the window, and lamp in the front-r | Anthropic: Claude Fable 5.1: Model B places all four elements correctly, inc…
Orbital Recall Grief
A close cinematic portrait of Commander Ilya Voss, a weathered human astronaut in a scuffed cobalt pressure underlayer, seated alone in the emergency greenhouse of the derelict research station Nacre-7, convincingly showing stunned grief just after hearing a final voice message from home—glassy eyes, trembling lower lip, clenched jaw failing to hold composure, one hand crushed around a wilted silver fern tag—ultra-detailed photorealistic sci-fi still, shallow depth of field, soft green grow-lights mixed with a cold rim light from a fractured porthole, asymmetric composition, 16:9
OpenAI: GPT-6 Luna: Model B conveys stunned grief more clearly with tearful eyes and a strained expression, and includes the wilted fern tag, derelict greenhouse, and fractured porthole. Model A better matches the requested close portrait and pressure suit, but its expression is less grief-stricken and it holds a fern rather than a tag. (Second judge pass, order swapped — scores are the average of both: Model B better matches the requested wide, seated greenhouse scene and clearly pairs the astronaut’s grief with t | Anthropic: Claude Fable 5.1: Model B nails nearly every brief element—seated…
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 B presents the teapot, glass, apple, and polished marble clearly, with convincing window reflections in the chrome and visible reflected color and light around the water. Model A has stronger dramatic side lighting, but the apple and its reflections are less clearly legible, especially in the glass. (Second judge pass, order swapped — scores are the average of both: Model B more clearly presents the chrome teapot, glass, apple, and polished dark marble, with strong window reflections in th | Anthropic: Claude Fable 5.1: Model B delivers a more convincingly photoreali…
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 requested lines correctly on a deep navy background. Model B’s stronger size hierarchy and more poster-like vertical spacing make the layout more polished, while Model A keeps the text compact and centered. (Second judge pass, order swapped — scores are the average of both: Model A keeps the three crisp, correctly spelled lines in a compact, centered stack with clear hierarchy, matching the prompt especially well. Model B has strong typography and a fitting portrait- | Anthropic: Claude Fable 5.1: Both render all three lines correctly spelled a…
Rain Bazaar Materials
A photorealistic street-level scene in the rain at the alloy bazaar of Kestra Quay on Europa: a prosthetics craftswoman fitting a polished brass finger to a dockworker under a patched indigo canvas awning, with believable wet metal reflections, swollen dark wood market tables, realistic skin translucency and pores, soaked wool sleeves, fogged glass display cases, and puddles catching distorted signage; physically accurate overcast daylight, soft shadowing, crisp material separation, documentary-style composition, 16:9
OpenAI: GPT-6 Luna: Model B more closely depicts a craftswoman fitting a polished brass finger in a rain-soaked bazaar, with convincing wet wood, puddle reflections, fogged display cases, and a patched indigo awning. Model A has appealing atmosphere but reverses the fitting interaction, shows a full prosthetic arm rather than a finger, and uses prominent black bars that weaken the composition. (Second judge pass, order swapped — scores are the average of both: Model B more closely matches the requested fitting of a | Anthropic: Claude Fable 5.1: Model A inverts the prompt by giving the crafts…
Neon Monsoon Midnight
A lone courier on a mag-lev scooter pauses beneath a tangle of hanging cables in the old neon ward of Sable Meridian during a monsoon midnight, the entire scene defined by electric cyan and magenta signage, rain haze, and deep violet shadows that create a moody, insomnia-soaked atmosphere; cinematic cyberpunk concept art, strong perspective down a narrow alley, reflective pavement, steam rising from vents, dramatic backlight and color spill, 16:9
OpenAI: GPT-6 Luna: Model B more clearly depicts a courier pausing on a scooter in a rain-soaked neon alley, with a delivery pack, hanging cables, steam, and strong depth. Model A has excellent atmospheric lighting and symmetry, but its rider and vehicle read less distinctly as a stopped courier scene. (Second judge pass, order swapped — scores are the average of both: Model B more clearly depicts a courier paused with a delivery load in a richly detailed neon alley, while Model A has a strong atmospheric compositi | Anthropic: Claude Fable 5.1: Model A delivers a stricter cyan/magenta/violet…
Matchup powered by OpenRouter.