FLUX.2 [dev] Flash vs GPT Image 1.5
Head-to-head across 7 challenges
FLUX.2 [dev] Flash
33.3%
win rate
Ties
11.1%
GPT Image 1.5
55.6%
win rate
Challenge Results
Geometric Composition
Text-to-Image“A glass cube on a wooden table. Inside the cube is a small blue sphere. On top of the cube sits a red book. A green plant is behind the cube, partially visible through the glass. Soft window light from the left.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Perfect adherence to the lighting instruction with a clear window light source on the left.
- + Highly realistic glass textures, including subtle imperfections and reflections.
- + Excellent depth of field and consistent perspective.
- − The sphere is slightly smaller than what might be expected from the prompt 'small', making it look a bit lost in the cube.
GPT Image 1.5
- + Excellent color saturation and vibrant subject matter.
- + Clean, sharp rendering of the glass cube and its internal reflection.
- + Good composition with the plant framing the background.
- − The lighting is a bit more diffuse and less directional than the requested 'window light from the left'.
- − The glass cube has a mirrored base which wasn't requested, creating a double reflection of the sphere.
Verdict: Both models followed the prompt instructions very well. FLUX.2 [dev] Flash stands out for its superior photographic realism and more accurate rendering of the soft directional window light. GPT Image 1.5 is also very strong but looks slightly more like a 3D render and added a mirrored floor to the cube that wasn't in the prompt.
Candid Street Photography
Text-to-Image“A candid street photo of an elderly Japanese man repairing a red bicycle in light rain, reflections on wet pavement, shallow depth of field, 50mm lens, natural skin texture, imperfect framing, motion blur from passing cars, cinematic but realistic, no stylization.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Excellent motion blur on background cars that creates a sense of movement.
- + Highly realistic skin texture and facial features of the elderly man.
- + Perfect adherence to 'imperfect framing' with a candid street photography feel.
- − The man's hands have structural issues, specifically the number and placement of fingers.
- − The bicycle kickstand and tools on the ground are somewhat messy and poorly defined.
GPT Image 1.5
- + Stronger emotional 'candid' feel with the man crouching and focused on his work.
- + Superior bicycle mechanics and tool kit detail.
- + Excellent depiction of rain droplets on the man's jacket and cap.
- − Failed to include 'motion blur from passing cars', as the car in the background is static.
- − The depth of field is slightly deeper than requested compared to the creamy bokeh in Model A.
Verdict: FLUX.2 [dev] Flash captures the cinematic technicalities better, specifically the motion blur and the 50mm lens look, but suffers from significant hand artifacts. GPT Image 1.5 provides a more grounded, realistic scene with better attention to the physical tools and the texture of rain, though it ignored the motion blur requirement. FLUX.2 [dev] Flash is the winner for its superior 'candid street' atmosphere and adherence to all photography-specific prompts despite the hand issues.
Fantasy Warrior
Text-to-Image“Close portrait of a battle-worn paladin in ornate engraved plate armor, hair braided with small beads, faint scars and dirt on the skin, warm torchlight reflecting off metal, shallow depth of field, bokeh sparks, lifelike eyes, highly detailed texture on leather straps and cloth underlayer.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Excellent adherence to the 'hair braided with small beads' detail
- + Clean and intricate engraving on the armor
- + Well-defined torchlight sources that balance the composition
- − The facial skin texture and blood effects look slightly flatter and less realistic than the competitor
- − The 'battle-worn' appearance feels a bit clean overall
GPT Image 1.5
- + Exceptional skin texture with realistic dirt and sweat mapping
- + Highly effective use of warm torchlight reflections on the metal
- + Strong sense of material depth in the leather and cloth underlayers
- − The beads in the hair are less distinct than requested
- − A small artifact is visible on the earring/ear area
Verdict: Both models followed the prompt exceptionally well, but GPT Image 1.5 wins due to its superior rendering of textures, particularly the skin and the interaction of light with the armor. While FLUX.2 [dev] Flash handled the specific request for hair beads more clearly, GPT Image 1.5 felt more like a tangible, high-quality photograph with a more convincing 'battle-worn' atmosphere.
Adorable Baby Animals in Sunny Meadow
Text-to-Image“Hyper-photorealistic scene of fluffy baby animals—a golden retriever puppy, tabby kitten, baby bunny, and red fox kit—with big expressive eyes and ultra-detailed soft fur, playfully chasing butterflies and tumbling together in a lush wildflower meadow, warm golden sunrise light with god rays and dew sparkles, joyful wholesome vibe, 8K masterpiece.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Excellent variety of animals that match the specific prompt ingredients.
- + High clarity and realistic lighting with clear god rays.
- + Balanced composition that fills the frame well without feeling cluttered.
- − Includes an extra fifth animal (a second rabbit/hare) not requested in the prompt.
- − Butterflies feel a bit static and repetitious in their placement.
GPT Image 1.5
- + Perfect adherence to the animal count and types requested.
- + Great dynamic action with the kitten in a 'tumbling' pose.
- + Strong rendering of 'warm golden sunrise light' and dew sparkles on the flowers.
- − The fox's eyes and snout look slightly more 'plush toy' than hyper-photorealistic.
- − The puppy's front paw has slightly odd toe definition.
Verdict: FLUX.2 [dev] Flash produces a cleaner, more technically impressive image with superior lighting, but it fails the prompt adherence check by adding a fifth animal. GPT Image 1.5 captures the 'tumbling' energy of the prompt much better and follows the numerical instructions exactly, making it the more reliable choice for this specific request.
Heroic Super Hero Portrait
Text-to-Image“Hyper-photorealistic full-body portrait of a female superhero standing triumphantly on a New York skyscraper rooftop at golden sunset, wearing a classic modest superhero costume with flowing cape, chest emblem, gloves, and boots in red and blue colors, practical design, short hair, strong determined heroic expression looking into the distance, powerful confident stance with hands on hips and cape billowing dramatically in the wind, detailed urban cityscape background, warm natural sunlight with sharp shadows and fabric highlights, ultra-sharp textures on suit, hair, and concrete, 8K masterpiece, empowering family-friendly style.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Excellent full-body composition with high level of photorealism
- + Realistic fabric textures and lighting integration on the character
- + Perfect adherence to the 'modest' attire and 'short hair' descriptors
- − The transition from the rooftop to the background cityscape looks slightly flat
- − The boot/leg anatomy looks a bit stiff
GPT Image 1.5
- + Strong, vibrant colors that pop against the sunset background
- + Dynamic cape physics and detailed facial construction
- + Very clear cityscape with recognizable landmarks
- − Failed the modesty requirement by featuring a very short skirt instead of a full suit
- − Light source on the character's face doesn't perfectly match the sun position behind her
- − The boots have a slightly plastic, less photorealistic appearance
Verdict: FLUX.2 [dev] Flash adhered much better to the specific constraints of the prompt, particularly the 'modest' costume requirement, providing a sleek full-body suit that feels practical. GPT Image 1.5 followed the aesthetic and backdrop well but ignored the modesty instruction by generating a micro-skirt, and its lighting feels more like a studio setup than natural sunlight.
Intricate Floral Mandala
Text-to-Image“Perfectly symmetrical mandala made entirely of real flowers, petals, leaves, fruits, and seeds in vibrant natural colors, intricate layered patterns with radial symmetry, top-down view on a soft neutral background, hyper-detailed organic textures and subtle shadows, photorealistic, 8K masterpiece.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Excellent variety of textures including soft petals, grainy fruit skins, and bumpy berries.
- + Sophisticated lighting with soft shadows that create a realistic sense of depth.
- + Highly photorealistic rendering of organic materials.
- − The radial symmetry is slightly imperfect, especially in the arrangement of the small berries and clusters.
- − Overall composition is a bit tighter/compressed compared to the other model.
GPT Image 1.5
- + Near-perfect mathematical radial symmetry across all layers.
- + High degree of clarity and sharpness for individual elements like seeds and acorns.
- + Great adherence to the 'mandala' geometric structure.
- − Lighting is a bit flat, making the objects look slightly more like a 2D digital illustration than a 3D photograph.
- − Texture on some leaves appears repetitive.
Verdict: FLUX.2 [dev] Flash produces a more believable photograph with superior organic textures and realistic lighting, though its symmetry is a bit organic and loose. GPT Image 1.5 excels at the technical mandala structure with perfect symmetry, but it feels less like a real physical object on a surface and more like a high-end graphic. FLUX.2 [dev] Flash is the winner for its mastery of 'photorealistic' material rendering.
Apollo 11: Journey to Tranquility
Text-to-Image“Create a clean, modern vector infographic poster about the Apollo 11 mission. NASA-inspired palette (navy, white, muted red, light gray). Flat-vector style, crisp lines, consistent iconography, subtle gradients only. Steps (stop at landing): 1. Launch (Saturn Vicon) 2. Earth Orbit (Earth + orbit ring icon) 3. Translunar (trajectory arc icon) 4. Lunar Orbit (Moon + orbit ring icon) 5. Descent (lunar module descending icon) 6. Landing (lunar module on the surface icon) Small supporting elements (minimal text): • Crew strip: three silhouette icons with only last names: Armstrong, Aldrin, Collins. • Landing site marker: Moon pin labeled "Tranquility" only. Layout constraints: generous margins, large readable labels, clean background with subtle stars. Vector-only, print-poster look, high resolution.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Includes all requested steps and icons with clear labels.
- + Higher details in the Lunar Module icons.
- + Accurately renders the names of all three astronauts.
- − Composition is cluttered and disorganized.
- − Contains redundant labels and hallucinatory text (e.g., 'Sataurr' Iccòn').
- − Style leans towards 'illustrated' rather than the requested 'clean flat-vector'.
GPT Image 1.5
- + Excellent adherence to the 'flat-vector' and 'clean infographic' aesthetic.
- + Superior layout with logical flow and consistent iconography.
- + Stronger adherence to the specified NASA-inspired color palette.
- − The 'Apollo 11' title is partially cut off at the top.
- − The translunar icon is simplified compared to the complex trajectory requested.
- − Includes some minor alignment issues between labels and icons.
Verdict: GPT Image 1.5 is the winner because it successfully captures the 'clean, modern vector infographic' aesthetic requested, using a clear vertical flow and consistent icons. While FLUX.2 [dev] Flash included more literal details, its composition is messy, redundant, and contains significant text hallucinations that detract from the professional poster feel.
FLUX.2 [dev] Flash
Fast distilled version of Black Forest Labs' FLUX.2 [dev] optimized for speed and cost efficiency.
GPT Image 1.5
OpenAI's state-of-the-art image generation model with better instruction following and adherence to prompts