Head to head: OpenAI: GPT-5.4 Image 2 vs Google: Nano Banana Pro (Gemini 3 Pro Image)
OpenAI: GPT-5.4 Image 2 vs Google: Nano Banana Pro (Gemini 3 Pro Image)
By Ryan Merket · Published
A head-to-head across eight image tasks tests style control, composition, anatomy, reflections and prompt-following. The key dividing line is whether a striking image also sticks closely to the requested framing and details.
Google’s Nano Banana Pro takes the saved result, winning **7 of 8 tasks**, with no ties. OpenAI’s GPT-5.4 Image 2 wins only **Cliffside mail leap**, where its airborne rider and dramatic action make the stronger read. The aggregate scores are **64.2 to 60.5** for Google; the editorial strength is **decisive**, not statistical. The pattern is clearest in the composition-heavy briefs. On the restricted-palette campsite, moonlit funicular poster and salt-flats observatory road, Google better follows the requested widescreen framing and keeps the key elements legible. It also wins the negation task by preserving the cozy reading-nook brief without adding prohibited objects. OpenAI’s images could be more atmospheric or richly detailed, but that polish did not consistently compensate when the output departed from the brief. In several tasks—including the campsite, poster, road and reading nook—its square framing was a specific mismatch; Google’s wins in hands and anatomy, reflections and glass, and the harbor-market still life round out its lead. **Verdict: Google’s Nano Banana Pro is the pick on this matchup, chiefly for more dependable adherence to composition and format.** That is an assessment of the saved task sample, not a claim about performance beyond it.
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 closely follows the four-color flat-vector brief and delivers a balanced 16:9 campsite scene with clear dusk atmosphere. Model A has appealing imagery, but its square format and visible gradients make it less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model B better matches the requested 16:9 flat-vector campsite scene, with a balanced arrangement of tents, campers, fire, lake, and mountains in the restricted palette. Model A has a polishe | Anthropic: Claude Fable 5.1: Model A keeps a clean four-color palette and fl…
Moonlit funicular poster
A clean vector travel poster of the hilltop town of Monte Serico at blue hour, showing a cream-colored funicular climbing between terraced cypress gardens toward a tiny domed belvedere under a huge rising moon, mood of quiet romance and anticipation, simplified geometric shapes, crisp outlines, limited palette of indigo, cream, muted coral, and pine green, elegant negative space, screen-printed illustration aesthetic, 16:9
OpenAI: GPT-6 Luna: Model B more closely matches the clean, geometric screen-printed poster style and requested 16:9 format, with the moon, cypress terraces, funicular, and belvedere arranged clearly. Model A is richly atmospheric and detailed, but its square format and busier, more painterly treatment are less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model B more closely matches the requested 16:9 clean vector poster, with crisp geometric shapes, a huge moon, terr | Anthropic: Claude Fable 5.1: Model B honors the requested 16:9 format, simpl…
Salt flats observatory road
A wide-angle architectural travel scene at dawn on the Salar de Arizaro salt flats: a straight raised road lined with tiny amber guide lamps leads toward a six-sided glass observatory and a distant row of blue service sheds, with two parked expedition trucks in the foreground and a radio mast farther back, all objects scaled believably smaller with distance and aligned to one consistent vanishing point, ultra-precise perspective, dry pastel sky reflected in the salt crust, clean high-detail realist rendering, 16:9
OpenAI: GPT-6 Luna: Model B better matches the requested widescreen composition and presents a clear straight, lamp-lined road converging on the observatory, with the sheds, trucks, and mast all visible. Model A has a more distinctive polygonal observatory and detailed expedition trucks, but its road bends noticeably and the square framing departs from the prompt. (Second judge pass, order swapped — scores are the average of both: Model B more closely matches the requested wide 16:9 scene, with the raised lamp-line | Anthropic: Claude Fable 5.1: Model B delivers the requested 16:9 frame, a dr…
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 shows a clear, plausible bracelet-making grip with soft light, but its square crop misses the requested 16:9 format. Model B better matches the wide candid-photo composition and natural setting while keeping both hands and the threads clearly visible. (Second judge pass, order swapped — scores are the average of both: Model A more clearly shows both hands in a plausible grip actively working on a recognizable friendship bracelet, while Model B has a stronger candid 16:9 setting but the b | Anthropic: Claude Fable 5.1: Model A renders clean, anatomically convincing…
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 follows the requested 16:9 framing and clearly shows the teapot, water glass, apple, and reflective surfaces, though the apple’s reflection in the chrome is not especially clear. Model A has striking reflections and detail, but its square framing departs from the prompt and the required reflections are also not fully evident in both surfaces. (Second judge pass, order swapped — scores are the average of both: Model B better honors the requested 16:9 close-up and gives the scene a strong, | Anthropic: Claude Fable 5.1: Model A clearly reflects the apple and window i…
Harbor market binding test
A sunlit quayside market stall in the fishing village of Vardhavn, arranged as a still-life travel scene: a matte cobalt-blue ceramic teapot, a glossy saffron-yellow glass buoy, a brushed copper lantern, a folded emerald-green waxed canvas raincoat, a pale lilac linen map pouch, and a charcoal-black cast-iron compass, each object clearly separate and keeping its own exact color and material with no color swapping or bleeding, soft morning side light, shallow depth of field, editorial product photography style, 16:9
OpenAI: GPT-6 Luna: Model A offers a richly detailed harbor-market still life with especially clear materials and colors, but its square framing misses the requested 16:9 format. Model B matches the wide format and includes all six distinct objects in warm harbor-side light, though the arrangement is less intimate and has some distracting background space. (Second judge pass, order swapped — scores are the average of both: Model B presents all six requested objects as a clear, polished still life in the specified w | Anthropic: Claude Fable 5.1: Model B delivers the requested 16:9 frame, genu…
Cliffside mail leap
A rainy coastal mountain road in the Faroe-style North Atlantic, captured in a cinematic freeze-frame as a scarlet vintage postal motorcycle launches over a broken section of cliffside pavement, the rider half-standing on the pegs with two canvas mailbags whipping sideways, spray and gravel arcing behind the tires, a startled flock of puffins exploding upward from the black basalt below, dramatic storm light breaking through slate clouds, telephoto action photography look with crisp subject motion and natural background blur, 16:9
OpenAI: GPT-6 Luna: Model B more clearly captures the requested 16:9 cliffside road leap, with the rider poised over a broken pavement edge, mailbags flying, and gravel scattering in cinematic storm light. Model A has a dramatic airborne pose and striking puffins, but its square framing and less clearly defined road leap make it a weaker fit overall. (Second judge pass, order swapped — scores are the average of both: Model A more convincingly captures the dramatic airborne leap, the rider braced on the motorcycle, | Anthropic: Claude Fable 5.1: Model A delivers the most dramatic read of the p…
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 has a cozy, polished reading nook and honors the exclusions, but its square framing misses the requested 16:9 format. Model B better matches the wide composition while including the chair, books, steaming tea, window, and warm afternoon light without any prohibited items. (Second judge pass, order swapped — scores are the average of both: Model B more closely matches the requested 16:9 framing while showing the cozy chair, books, tea, and warm window light without any prohibited objects. | Anthropic: Claude Fable 5.1: Both Model A and Model B correctly omit plants,…
Matchup powered by OpenRouter.