Black Forest Labs' open-weights image generation model with frontier performance, available for non-commercial local deployment
Settled by community votes across 13 shared challenges, with an AI judge weighing in on each.
FLUX.2 [dev]
#18 of 62 in Text-to-Image
Stable Diffusion 3.5 Large
#30 of 62 in Text-to-Image
Where the votes landed
FLUX.2 [dev]
75.0%
win rate
Ties
0.0%
Stable Diffusion 3.5 Large
25.0%
win rate
Challenge by challenge
The strongest take from each model on every shared challenge, with the AI judge's read.
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]
- + Perfect adherence to all spatial instructions
- + Realistic glass physics and refractions
- + High-quality photographic texture on the book and table
Stable Diffusion 3.5 Large
- + Crisp, sharp details on the glass edges
- + Vivid colors
- − Failed the spatial reasoning: the book is inside the cube instead of on top
- − The blue sphere appears to be floating unnaturally above the book
- − The plant is mostly above/beside the cube rather than behind it
Verdict: FLUX.2 [dev] followed every instruction perfectly, correctly placing the sphere inside the cube and the book on top. Stable Diffusion 3.5 Large failed the spatial arrangement, placing the book inside the cube and the sphere on top of the book, which was the opposite of what was requested.
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]
- + Excellent motion blur from passing cars as requested
- + Highly realistic skin texture and facial features
- + Superior pavement reflections and wet weather atmosphere
- − The bicycle structure is mechanically confusing with overlapping wires and frames
- − The framing is a bit centered despite the 'imperfect framing' prompt
Stable Diffusion 3.5 Large
- + Accurate bicycle anatomy and silhouette
- + Effective shallow depth of field
- + Good color contrast between the red bike and the teal-toned background
- − Lacks the requested motion blur on the vehicles
- − The rain effect looks like a static overlay rather than a natural part of the scene
- − Skin texture is slightly smoothed and less 'natural' than Model A
Verdict: FLUX.2 [dev] significantly outperformed Stable Diffusion 3.5 Large in capturing the 'cinematic but realistic' atmosphere, specifically regarding the complex motion blur of passing traffic and the gritty realism of the subject's skin. While Stable Diffusion 3.5 Large produced a cleaner bicycle, it failed to incorporate the motion blur requested and the rain looked less integrated into the environmental lighting.
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]
- + Excellent adherence to the 'beads' requirement in the hair braids.
- + Very realistic lighting with clear torchlight reflections on the metal and skin.
- + Highly detailed and legible textures on the leather straps and metal engravings.
- − The facial scars look a bit like fresh paint or surface marks rather than deep, healed scars.
- − The hair logic where it transitions into thin braids is slightly messy.
Stable Diffusion 3.5 Large
- + Incredible detail on the chainmail and cloth underlayer texture.
- + The engraving on the armor is more intricate and fine-grained.
- + Very realistic facial skin texture and natural weathering effects.
- − Completely missed the requirement for 'small beads' in the hair.
- − The lighting is flatter and lacks the strong warm 'torchlight' glow requested.
- − The bokeh sparks are barely visible compared to the other model.
Verdict: FLUX.2 [dev] followed the prompt more comprehensively, specifically including the requested hair beads and creating a much stronger atmosphere with warm torchlight and visible bokeh sparks. While Stable Diffusion 3.5 Large has superior fine-detail textures in the armor and cloth, it failed a key descriptor (beads) and lacks the requested lighting drama.
Modern Clean Menu
Text-to-Image“Modern minimalist restaurant menu design, white background with colorful food photos in grid, sections for appetizers/pizza/mains, bold sans-serif fonts, vibrant accents, clean professional layout for casual dining.”
AI Judge Analysis
FLUX.2 [dev]
- + Excellent structural layout that accurately resembles a functional restaurant menu.
- + Clear section headers for Appetizers, Pizza, and Mains as requested.
- + Realistic food photography with appetizing lighting.
- − Text contains several gibberish words and spelling errors (e.g., 'PIZZAU').
- − The colorful accents are a bit repetitive and chunky.
Stable Diffusion 3.5 Large
- + Stronger adherence to the 'grid' aspect for the food photos.
- + Very bold, modern sans-serif typography that feels high-end.
- + Clean white background with excellent negative space.
- − The menu layout is less practical, feeling more like a poster than a multi-section menu.
- − Text legibility is low for the smaller items with many symbols/artifacts.
- − Mains section (MAIMAES) is misspelled.
Verdict: FLUX.2 [dev] produces a much more realistic and usable menu layout with clear price points and distinct sections, though its text rendering is imperfect. Stable Diffusion 3.5 Large has a more artistic, high-fashion aesthetic but fails on functional layout and basic spelling for the requested headers. FLUX.2 [dev] is preferred for its better interpretation of a 'professional layout for casual dining'.
Magic Burger Explosion: Fiery Photorealism Challenge
Text-to-Image“Ad for 'Magic Burger'. Dynamic, exploded burger with all components (bun, patty, cheese, lettuce, tomato, sauce) suspended in mid-air. Emphasize photorealistic detail and a sense of motion. Dark, fiery background with glowing embers. Integrate text: 'MAGIC BURGER' as a prominent title, 'LIMITED TIME ONLY' as a secondary message, and '€6.99' in a starburst, all rendered with a fiery, glowing effect.”
AI Judge Analysis
FLUX.2 [dev]
- + Excellent text integration with flawless spelling and fiery glow effects.
- + Perfect adherence to the 'exploded' layout requested in the prompt.
- + Professional composition suitable for a commercial advertisement.
- − The burger ingredients look slightly more synthetic compared to the textures in Model B.
- − The lighting on the bottom bun is a bit flat.
Stable Diffusion 3.5 Large
- + High-quality photorealistic texture on the meat and charcoal.
- + Dynamic fire and ember effects integrated into the burger itself.
- − Completely failed to include any of the requested text elements.
- − Did not follow the instruction for an 'exploded' view with suspended components.
- − The burger is stacked rather than suspended in mid-air.
Verdict: FLUX.2 [dev] followed every single instruction within the prompt, including complex text rendering and a specific 'exploded' layout. In contrast, Stable Diffusion 3.5 Large ignored all text requirements and the specific exploded composition, offering a standard stacked burger instead. FLUX.2 [dev] is the clear winner for its superior prompt adherence and layout execution.
Chalkboard Menu
Text-to-Image“Handwritten-style chalkboard menu in a cozy café, all text rendered in the exact same realistic chalk handwriting style with natural variations in letter size, slight slant, and chalk texture — no printed or digital fonts anywhere on the board. Title at the top in elegant cursive chalk handwriting: ‘TODAY’S SPECIALS – APRIL 30, 2026’. Below it, three menu items also in the same handwritten chalk style: ‘Truffle Mushroom Risotto – $24’, ‘Grilled Octopus with Lemon & Herbs – $28’, ‘Brown Butter Chocolate Chip Cookies – $9’. At the very bottom, smaller text in the identical handwritten chalk style (slightly smaller but still clearly legible with the same handwriting characteristics): ‘All items made fresh daily • Ask about our gluten-free options’. Warm ambient café lighting, visible chalk dust and smudges, realistic handwriting imperfections, no clean printed text anywhere.”
AI Judge Analysis
FLUX.2 [dev]
- + Excellent text rendering with near-perfect spelling and handwriting style.
- + Strong adherence to the specific date and menu items requested.
- + Realistic chalk texture including smudges and dust on the board.
- − The layout is a bit tight with text reaching the very edges of the board.
Stable Diffusion 3.5 Large
- + Features a beautiful, wide-angle composition of the café environment.
- + Follows the general aesthetic of a chalkboard menu board.
- − Numerous spelling errors including 'TODAAY' and 'Cholcalte Chip'.
- − Failed to follow the requested year, displaying 2024 instead of 2026.
- − The text is very messy and difficult to read compared to the other model.
Verdict: FLUX.2 [dev] significantly outperformed Stable Diffusion 3.5 Large by accurately rendering almost all requested text with realistic chalk textures and correct spelling. While Stable Diffusion 3.5 Large provided a more comprehensive environmental shot of the café, it failed on the specific menu items, the date, and basic legibility.
The Reversed Rodeo
Text-to-Image“Horse riding astronaut in space — horse on top, not vice versa. Surreal, highly detailed, cinematic.”
AI Judge Analysis
FLUX.2 [dev]
- + Excellent photorealistic lighting and textures on both the space suit and the horse.
- + Very clean composition with clear separation between foreground and background assets.
- + High level of detail in the horse's anatomy and the mechanical parts of the astronaut's pack.
- − Failed the specific spatial instruction 'horse on top, not vice versa'.
- − The shadow on the clouds below does not match the horse's pose.
Stable Diffusion 3.5 Large
- + Beautiful cinematic atmosphere with a gritty, star-flecked texture.
- + Dynamic sense of motion with the horse galloping through cosmic dust.
- + The scale of the planet in the background adds a grand, surreal feel.
- − Failed the specific spatial instruction 'horse on top, not vice versa'.
- − The horse's face/muzzle area has slightly muddy details compared to Model A.
Verdict: Both models failed the negative constraint to have the horse on top of the astronaut, instead defaulting to the common trope of an astronaut riding a horse. FLUX.2 [dev] produces a much cleaner, more technically proficient image with superior lighting, while Stable Diffusion 3.5 Large offers a more artistic and ethereal 'cinematic' style that feels more surreal.
The Capybara Taxi Driver
Text-to-Image“Photorealistic scene inside a yellow New York taxi at night. A capybara is driving, wearing a yellow taxi driver cap and a dark jacket. It has a calm, professional expression and both front paws on the steering wheel. In the back seat sits a human businesswoman in a coat, looking at her phone with a completely normal, bored expression (as if this is just another normal ride). Through the windows you can see the streets of Manhattan at night with blurred lights. Realistic taxi interior, photorealistic, detailed fur and fabric, 35mm lens, night lighting with reflections, shallow depth of field.”
AI Judge Analysis
FLUX.2 [dev]
- + Excellent adherence to the full prompt, including the passenger and lighting.
- + Highly realistic textures for both the capybara's fur and the passenger's clothing.
- + Captures the requested 'bored' expression on the businesswoman perfectly.
- − The capybara's paws look slightly human-like in their structure.
Stable Diffusion 3.5 Large
- + Strong colors and sharp details on the capybara's clothing.
- + Creative and high-quality rendering of the capybara's face and whiskers.
- − Completely missed the businesswoman in the back seat.
- − The capybara only has one hand on the wheel instead of two as requested.
- − The perspective makes the capybara look unnaturally small or positioned poorly in the seat.
Verdict: FLUX.2 [dev] successfully followed every detail of the complex prompt, creating a coherent narrative scene with both the driver and the bored passenger. Stable Diffusion 3.5 Large failed to include the passenger entirely and missed the specific instruction for both paws to be on the steering wheel, focusing only on the driver.
The Halloween Invitation
Text-to-Image“Vintage gothic Halloween party invitation. Dark parchment poster, spooky border with webs and thorns, central glowing jack-o-lantern, bats, twisted trees, moody night sky. Add elegant gothic title text saying "Halloween Party Invitation", a small scroll banner saying "You are invited to a night of frights", and event details at the bottom: Date: 30.10.2026 Time: 7pm Location: The Arches, NYC Spooky but polished, cinematic lighting, square format.”
AI Judge Analysis
FLUX.2 [dev]
- + Perfect text rendering for all requested details, including date and location.
- + High-quality gothic aesthetic with a cohesive thorn and web border.
- + Excellent layout balance that follows all prompt instructions precisely.
- − The lighting in the background sky is a bit flat compared to the pumpkin.
Stable Diffusion 3.5 Large
- + Wonderful vintage parchment texture with burnt/torn edges.
- + Dynamic composition with a large moon and twisted tree silhouettes.
- + Good atmospheric lighting.
- − Failed to include the event details (date, time, location) at the bottom.
- − Text rendering on the scroll is messy and looks like handwriting rather than elegant gothic print.
- − The layout feels cramped at the bottom with the missing information.
Verdict: FLUX.2 [dev] is the clear winner as it successfully rendered every piece of text requested, including the specific date and location which Stable Diffusion 3.5 Large ignored completely. FLUX.2 [dev] also provided a more polished and professional 'invitation' feel with a superior thorn border.
Isometric Miniature Diorama Scenes
Text-to-Image“Create a clear, 45° top-down isometric miniature 3D cartoon scene of Japan's signature dish: sushi, with soft refined textures, realistic PBR materials, gentle lighting, on a small raised diorama base with minimal garnish and plate. Solid light blue background. At top-center: 'JAPAN' in large bold text, 'SUSHI' below it, small flag icon. Perfectly centered, ultra-clean, high-clarity, square format.”
AI Judge Analysis
FLUX.2 [dev]
- + Excellent adherence to the 45° isometric diorama layout
- + Perfect rendering of text and flag icon exactly as requested
- + Clean, soft, minimal 3D cartoon aesthetic
- − The sushi variety is slightly limited compared to the competitor's detail
Stable Diffusion 3.5 Large
- + Highly realistic and appetizing textures on the fish and rice
- + Complex and detailed compositions with vibrant colors
- − Failed the text placement and background requirement
- − The perspective is more of a standard 3/4 view rather than a true isometric diorama
- − Added unrequested objects like chopsticks and a small bowl
Verdict: FLUX.2 [dev] followed every instruction perfectly, including the specific isometric diorama format, text layout, and minimal background requirements. Stable Diffusion 3.5 Large produced more realistic fish textures but ignored the layout and text constraints, resulting in a cluttered scene that didn't meet the prompt's structural goals.
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]
- + Excellent fur texture and realistic anatomy for all four animals.
- + Superior lighting with clear god rays and convincing dew sparkles on the grass.
- + Very high level of detail in the foreground flowers and animal eyes.
- − The animals are mostly sitting still rather than 'playfully chasing' and 'tumbling' as requested.
- − Generated two bunnies instead of just one.
Stable Diffusion 3.5 Large
- + Captures the 'playfully chasing' and 'tumbling' movement much better than the other image.
- + Includes one of each requested animal as specified in the count.
- + Strong bokeh effect creates a whimsical, commercial look.
- − The anatomy of the animals is slightly distorted, particularly the puppy's paws and the fox's face.
- − The 'kitten' looks more like a small fox or a hybrid, lacking clear tabby markings.
- − Lower overall texture detail compared to the competitor.
Verdict: FLUX.2 [dev] produces a much more realistic and detailed image with beautiful lighting and textures, though it fails on the specific action of 'tumbling' and adds an extra bunny. Stable Diffusion 3.5 Large captures the dynamic energy of the prompt much better, but suffers from anatomical inconsistencies and less realistic fur rendering. FLUX.2 [dev] is the preferred choice for its sheer visual quality and adherence to the 'hyper-photorealistic' part of the prompt.
Vintage Cafe Logo
Text-to-Image“Vintage minimalist restaurant logo for "Caffè Florian", retro cloche dome with steam and "Est. 1720" banner, classic typography, warm brown and cream tones, subtle texture on light background, vector emblem style.”
AI Judge Analysis
FLUX.2 [dev]
- + Excellent typography with correct spelling and accent placement.
- + Professional vector emblem style with clean, balanced lines.
- + Effective use of subtle paper texture on the background.
- − The steam icon is a bit small and generic.
Stable Diffusion 3.5 Large
- + Creative use of negative space for the steam effect.
- + Incorporates decorative corner elements for a vintage feel.
- + Clear resolution and adherence to the warm color palette.
- − Includes a spelling error in 'Cafféé'.
- − The cloche dome is floating and lacks structural coherence with its base.
- − The 'Est. 1720' text is outside of the banner.
Verdict: FLUX.2 [dev] followed the prompt more accurately, producing a cohesive vector emblem with perfect typography. Stable Diffusion 3.5 Large had a spelling error and failed to place the 'Est. 1720' text within the requested banner, though it did offer a creative take on the steam effect.
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]
- + Excellent adherence to the iconography requested for each step.
- + Legible typography for major headers and some labels.
- + Clean, professional vector aesthetic that matches the 'modern infographic' request.
- − Text labels are disorganized/redundant (e.g., repeating 'Saturn Icon' instead of just 'Launch').
- − The sequence is laid out in an unintuitive grid that doesn't flow clearly from 1 to 6.
- − Some minor spelling errors in labels like 'Tranquility' and 'Liaonides'.
Stable Diffusion 3.5 Large
- + Strong composition with a cohesive, complex technical layout.
- + Good use of the NASA-inspired color palette.
- + Includes a large, detailed lunar graphic that grounds the poster.
- − Text is largely unintelligible gibberish.
- − The rocket shown is a generic shuttle-style craft rather than a Saturn V.
- − Failed to create a clear step-by-step numbered/ordered infographic as requested.
Verdict: FLUX.2 [dev] followed the prompt instructions much more closely, providing the specific icon types and mission steps requested, even if the layout order is a bit jumbled. Stable Diffusion 3.5 Large produced a visually interesting poster but failed to render a Saturn V rocket and completely ignored the structured 6-step infographic requirement, opting instead for a decorative but non-functional layout with garbled text.
Explore each model
Stability AI's 8.1-billion parameter Multimodal Diffusion Transformer (MMDiT) text-to-image model featuring improved image quality, typography, complex prompt understanding, and resource-efficiency