Head to head: FLUX 2 vs Ideogram V4.5 Edit
FLUX 2 vs Ideogram V4.5 Edit
By Ryan Merket · Published
A head-to-head across eight image tasks tests text rendering, strict visual constraints, composition, and prompt-specific details. The results show different strengths in structured scenes and tightly specified aesthetics.
Ideogram V4.5 Edit takes the **clear editorial result** on this saved task sample, with an aggregate score of **62.9** to FLUX 2’s **60.0**. It won five tasks to FLUX 2’s two, with one tie—a spread built on several distinct prompt-following wins, not a sweep. Ideogram’s strongest showings came in the car-free silk district, Memphis-palette atelier, named art style, and restricted-palette tasks. It brought richer noir atmosphere and street detail, arranged the atelier objects more completely within the requested palette, made the ukiyo-e boat-and-wave relationship clearer, and stayed closer to the four-color flat-vector constraint. On multi-line text, both models rendered the lines correctly; the saved verdict favors Ideogram, though FLUX’s layout was judged more restrained and crisp. FLUX had the edge where object relationships and emotional expression were central: its blue cylinder was more clearly placed behind the foreground objects, and its backstage portrait better conveyed exhausted relief. The atrium trenchcoat-reflections task was a tie: FLUX made the cracked mirror more apparent, while Ideogram delivered a polished cinematic composition and stronger rain-soaked atmosphere. **Verdict: Ideogram V4.5 Edit wins this matchup, 62.9 to 60.0.** The assessment applies to this saved task sample; it is not a claim about performance beyond these tasks.
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 all three lines clearly and correctly in a balanced, minimalist layout with subtle grain. Model B also renders the text accurately, but its much larger headline and heavier texture feel less restrained than the prompt requests. (Second judge pass, order swapped — scores are the average of both: Both models render all three lines correctly on a deep navy background, but Model A has a more balanced, tightly centered layout and especially crisp typography. Model B’s title dominates | Anthropic: Claude Fable 5.1: Both render all three lines correctly, but Mode…
Atrium trenchcoat reflections
A hyperreal fashion editorial in a rain-glossed botanical conservatory at dusk, featuring a model in a smoked-quartz vinyl trench coat and mirrored ankle boots pausing beside a tall cracked antique mirror, with hanging ferns, brass railings, and a koi pond all visible through curved glass walls; the image must show accurate reflections in the mirror, believable refraction and distortion through the greenhouse panes, and coherent mirrored highlights on the chrome boot surfaces, composed as a wide cinematic shot with moody teal-gold lighting, 16:9.
OpenAI: GPT-6 Luna: Model A more clearly includes the cracked mirror and has vivid chrome-boot highlights, but Model B has a more polished cinematic composition, convincing rain-wet conservatory details, and strong teal-gold atmosphere. Model B’s main omission is that the antique mirror does not visibly appear cracked. (Second judge pass, order swapped — scores are the average of both: Model A more clearly delivers the cracked antique mirror, glossy vinyl coat, and coherent reflective boots in a polished cinematic | Anthropic: Claude Fable 5.1: Model A nails the specified materials and mood—a…
Car-free silk district
A film-noir black-and-white street scene in the old silk market quarter of Valmere just after rain, showing couture window displays, mannequins in angular eveningwear, neon reflections on cobblestones, a newspaper kiosk, and two pedestrians under one umbrella—while the avenue is completely empty of cars, taxis, buses, bicycles, scooters, and any other vehicles; shot as a low wide composition with dramatic storefront glow and deep shadows, 16:9.
OpenAI: GPT-6 Luna: Both images capture the rain-soaked, vehicle-free noir street with couture displays, a kiosk, and umbrella-carrying pedestrians. Model B has richer street detail and a more immersive wide composition, while Model A’s Valmere sign is a particularly clear prompt-specific touch. (Second judge pass, order swapped — scores are the average of both: Model B delivers the stronger noir atmosphere, with vivid wet-cobblestone reflections, dramatic storefront glow, couture displays, and a prominent newspape | Anthropic: Claude Fable 5.1: Model B delivers a richer film-noir atmosphere…
Memphis palette atelier
A crisp flat-lay fashion still life of an avant-garde accessory designer’s worktable arranged with a pleated scarf, resin bangles, oversized sunglasses, thread cards, sketchbook, tailor’s chalk, and cut fabric swatches, rendered strictly in the named Memphis palette only: turquoise, coral, lemon yellow, black, and white, with absolutely no off-palette colors anywhere; graphic editorial style with hard noon-like studio lighting, sharp shadows, and meticulous geometric composition, 16:9.
OpenAI: GPT-6 Luna: Model B includes the requested atelier objects in a more complete, meticulously arranged flat lay, with a strong geometric rhythm and a closer match to the restricted palette. Model A has especially vivid lighting and impact, but its bangles and small printed details introduce more noticeable off-palette colors. (Second judge pass, order swapped — scores are the average of both: Model B includes the full set of requested atelier objects in a particularly meticulous geometric flat-lay, with a pal | Anthropic: Claude Fable 5.1: Model B holds the Memphis palette far more stri…
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 directly depicts a fishing boat riding the cresting wave, with especially strong ukiyo-e foam, bold outlines, and a cohesive indigo-and-cream palette. Model A has an appealing dawn atmosphere and woodblock-print detail, but the boat is less clearly riding the wave and the composition is more crowded. (Second judge pass, order swapped — scores are the average of both: Model B more clearly shows a fishing boat riding the towering cresting wave, with bold outlines, stylized foam, and a | Anthropic: Claude Fable 5.1: Model B delivers a coherent ukiyo-e scene with…
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 A presents the requested object colors and relationships clearly, with the blue cylinder visibly behind both foreground objects and a balanced tabletop composition. Model B also includes the correct objects, but the cylinder is more obscured and the oversized cube and broad empty space make the arrangement less clear and balanced. (Second judge pass, order swapped — scores are the average of both: Model B follows the requested object relationships and color assignments, with soft studio li | Anthropic: Claude Fable 5.1: Model A satisfies every binding: matte green cu…
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 A has a more atmospheric dusk scene and richer composition, but its gradients and many tonal variations conflict with the strict four-color, no-gradient requirement. Model B is a cleaner flat-vector design that stays much closer to the restricted palette, though its composition is simpler and its bright sky reads less clearly as dusk. (Second judge pass, order swapped — scores are the average of both: Model B closely follows the four-color flat-vector constraint and presents a clear, balan | Anthropic: Claude Fable 5.1: Model A violates the core constraint with a cle…
Backstage relief close-up
An intimate editorial portrait of a young tailor backstage after salvaging a disastrous runway hemline, conveying unmistakable exhausted relief with damp eyes, a half-laughing smile, loosened shoulders, and trembling hands still holding silver dressmaker shears; styled as medium-format fashion photography with shallow depth of field, powder-blue work apron dusted with chalk, warm tungsten bulbs and one cool sidelight shaping the face, close three-quarter composition.
OpenAI: GPT-6 Luna: Model A better conveys the requested exhausted relief through the closed, damp-looking eyes and half-laughing smile, with a close composition, chalky blue apron, shears, and contrasting warm and cool light. Model B has a polished backstage setting but reads as cheerfully posing rather than visibly exhausted or trembling, and its framing is less intimate. (Second judge pass, order swapped — scores are the average of both: Model A better captures the requested exhausted relief through damp-looking | Anthropic: Claude Fable 5.1: Model A delivers a close three-quarter portrait…
Matchup powered by OpenRouter.