FP8 quantized variant of Black Forest Labs' FLUX.1 [schnell] model, offering ~2x faster inference with reduced precision while maintaining high-quality image generation in 4 steps
Settled by community votes across 13 shared challenges, with an AI judge weighing in on each.
FLUX.1 [schnell] FP8
19.0
arena score
#47 of 62 in Text-to-Image
Skill signature
· Text-to-Image
Stable Diffusion 3.5 Large Turbo
10.7
arena score
#61 of 62 in Text-to-Image
Vote tally
Where the votes landed
FLUX.1 [schnell] FP8
0%
win rate
Ties
0%
Stable Diffusion 3.5 Large Turbo
0%
win rate
Shared challenges
13
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Excellent material rendering with realistic glass thickness and light refraction.
- + Very high visual quality with soft, natural window lighting.
- + Correctly placed the red book on top of the glass cube.
- − The glass object is a tall rectangular prism rather than a perfect cube.
- − Includes an extra internal glass shelf that was not requested.
Stable Diffusion 3.5 Large Turbo
- + Correctly depicted a cube shape with distinct black framing.
- + Accurately positioned the green plant behind the cube as requested.
- − Failed the spatial instruction by placing the sphere on the book and the book inside the cube.
- − The lighting is harsh and the materials look more like 3D renders than a photograph.
Verdict: FLUX.1 [schnell] FP8 produced a much more realistic image with superior lighting and material properties, successfully placing the red book on top of the glass. While Stable Diffusion 3.5 Large Turbo followed the geometric 'cube' instruction better, it failed the core spatial logic by placing the book inside the cube instead of underneath the sphere.
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Excellent adherence to the 'cinematic' and 'wet pavement reflection' prompts.
- + Realistic skin textures and complex lighting interaction.
- + Natural-looking background with logical depth of field and Japanese signage.
- − Lack of visible 'motion blur' on the passing cars as requested.
- − The man's hands are slightly merged with the bicycle handlebar area.
Stable Diffusion 3.5 Large Turbo
- + Stronger adherence to the 'light rain' prompt with visible falling droplets.
- + Good portrayal of the man's posture for the task of repairing.
- − Low visual quality with significant anatomical errors in the face and hands.
- − The environment feels sterile and lacks the requested cinematic reflections and wet pavement detail.
- − The bicycle and the subject look like 3D assets pasted onto a background rather than integrated.
Verdict: FLUX.1 [schnell] FP8 is the clear winner as it produces a high-quality, atmospheric, and realistic street photograph that aligns with the aesthetic requirements of the prompt. In contrast, Stable Diffusion 3.5 Large Turbo suffers from significant distortion in the subject's face and hands, resulting in a much lower-quality image.
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Excellent play of warm torchlight with high dynamic range on the face.
- + Highly detailed and realistic skin texture with visible pores and subtle scarring.
- + Stronger interpretation of the 'battle-worn' aesthetic with intense, lifelike eyes.
Stable Diffusion 3.5 Large Turbo
- + Accurately followed the prompt for braided hair.
- + Ornate engraving on the plate armor is well-defined.
- + Clear separation of layers including the cloth underlayer.
- − Skin texture appears plastic or heavily airbrushed compared to the other model.
- − Lighting is flat and lacks the 'warm torchlight' atmosphere requested.
- − Blood/dirt on the face looks like digital paint rather than integrated texture.
Verdict: FLUX.1 [schnell] FP8 delivers a much more cinematic and realistic portrait with superior lighting and skin texture, though it missed the specific detail of braided hair. Stable Diffusion 3.5 Large Turbo followed the hair prompt better, but the overall image quality suffers from a plastic look and flat lighting that fails to capture the 'battle-worn' mood.
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Excellent structure that resembles a real double-sided menu.
- + Includes all requested sections: appetizers, pizza, and mains.
- + Professional use of grids and hierarchy.
- − Several spelling errors in headings ('Apptizers', 'Orceters', 'Seccer').
- − Small body text is mostly gibberish.
Stable Diffusion 3.5 Large Turbo
- + High saturation and vibrant food photography.
- + Good clarity on the main 'Pizza' and 'Mians' headers.
- + Creative top-down composition.
- − Poor layout for an actual menu, looking more like an art piece.
- − Incorrect spelling of 'Mains' as 'Mians'.
- − Missing the 'Appetizers' section header.
Verdict: FLUX.1 [schnell] FP8 followed the prompt much more accurately by creating a functional, grid-based menu layout with appropriate sections and pricing, despite some spelling errors. Stable Diffusion 3.5 Large Turbo produced more of a collage of food images and lists that don't coalesce into a professional document design.
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Successfully integrated all requested text elements, including the title and price.
- + Captures highly photorealistic textures on the bun and meat patty.
- + Includes the requested starburst element for the price.
- − Significant spelling errors in the secondary text ('LIIMITED', 'NEEY').
- − Failed to properly 'explode' the burger, keeping the main stack mostly intact instead of suspending components separately.
- − The price is incorrect (€69 vs €6.99).
Stable Diffusion 3.5 Large Turbo
- + Strong artistic interpretation of the 'fiery background' with glowing embers and smoke.
- + Better sense of vertical suspension and lighting interaction with the fire.
- − Completely failed to include any of the requested text ('MAGIC BURGER', price, etc.).
- − Image quality appears more like digital art than the requested 'photorealistic detail'.
- − Did not follow the 'exploded' instruction, showing a mostly assembled burger.
Verdict: Both models struggled with the 'exploded' layout, but FLUX.1 [schnell] FP8 is the clear winner for following the complex text integration portion of the prompt, despite spelling errors and the wrong price. Stable Diffusion 3.5 Large Turbo failed to include any text at all, which was a core requirement of the advertisement prompt.
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + The text has a much more authentic handheld chalk texture and natural handwriting style.
- + The board fills the frame well, creating a more intimate 'café' feel.
- + Spelling in the main header is accurate.
- − The menu items contain several spelling errors and repetitive lines ('Grillled Octemon').
- − Failed to render the heading in elegant cursive as requested.
Stable Diffusion 3.5 Large Turbo
- + The environment is clean and well-lit with attractive secondary elements like plants.
- + The title font is closer to a cursive/script style.
- + The layout is organized into neat columns.
- − The text looks like a digital font rather than natural chalk handwriting.
- − Severe hallucinations in the text, including dates ('apri 3l 20-') and prices.
- − The 'handwritten' request was ignored in favor of a stylized graphic design look.
Verdict: FLUX.1 [schnell] FP8 is the clear winner for its superior ability to mimic the texture and variability of actual chalk handwriting, despite some spelling issues. Stable Diffusion 3.5 Large Turbo produced an image that looks like a digital menu board with significant text gibberish, failing to capture the 'handwritten' and 'chalk texture' requirements of the prompt.
The Reversed Rodeo
Text-to-Image“Horse riding astronaut in space — horse on top, not vice versa. Surreal, highly detailed, cinematic.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Excellent adherence to the 'horse on top' spatial instruction
- + Cinematic lighting and high-quality textures on the Earth's surface
- + Creative interpretation of a horse carrying astronaut equipment
- − The second horse head appearing from behind is confusing and may be an anatomy glitch
- − No visible human astronaut, only equipment
Stable Diffusion 3.5 Large Turbo
- + Very clean and sharp rendering of the astronaut and horse
- + Good composition with the curvature of the atmosphere
- − Failed the negative constraint entirely by placing the astronaut on top
- − Lacks the 'surreal' quality requested, looking like a standard sci-fi image
Verdict: FLUX.1 [schnell] FP8 followed the specific instruction to have the horse on top of the astronaut (or astronaut equipment), correctly interpreting a difficult spatial prompt. Stable Diffusion 3.5 Large Turbo produced a high-quality image but failed the core challenge by generating a standard image of an astronaut riding a horse.
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Excellent character placement and framing that clearly shows both subjects.
- + Superior lighting consistency between the interior and the street lights.
- + Effective storytelling with the businesswoman holding two phones, appearing busy.
- − The capybara's head shape is slightly distorted to fit the forward-facing pose.
- − The cap is more of a toy hat than a realistic driver cap.
Stable Diffusion 3.5 Large Turbo
- + High-quality textures on the capybara's fur and whiskers.
- + Very realistic clothing and steering wheel details.
- + The taxi driver cap looks like a real piece of apparel.
- − The passenger in the back is heavily blurred and her phone is not visible.
- − Failed to include the 'looking at her phone' requirement for the human character.
Verdict: FLUX.1 [schnell] FP8 is the winner because it successfully captured all elements of the prompt, including the specific interaction and expression of the human passenger. Stable Diffusion 3.5 Large Turbo produced a very high-quality animal portrait but failed to properly render the secondary subject and her required actions.
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Successfully included all the specific text information from the prompt.
- + Excellent moody atmosphere with cinematic lighting and a dark parchment aesthetic.
- + Layout effectively follows the hierarchy of an invitation.
- − Several typos in the small text such as 'nigh tof' and 'Locaation'.
- − The bottom background is a bit too dark, making the tree silhouettes blend into the abyss.
Stable Diffusion 3.5 Large Turbo
- + Clean, graphic-novel style with very clear spiderweb and thorn border details.
- + High contrast and sharp rendering of the central jack-o-lantern.
- − Failed to include the specific banner text, date, time, and location entirely.
- − Missed the 'dark parchment' and 'gothic title' aesthetic, opting for a generic font and light background.
- − Art style feels more like clip-art than a vintage gothic invitation.
Verdict: FLUX.1 [schnell] FP8 is the clear winner as it followed all prompt instructions, including the specific event details, and captured the requested 'vintage gothic' mood perfectly. Stable Diffusion 3.5 Large Turbo completely ignored the event details (date, time, location) and failed to include the scroll banner, resulting in a generic image rather than a functional invitation.
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Excellent soft refined 3D textures that match the 'cartoon scene' vibe well.
- + Highly accurate 45-degree isometric perspective and clean centering.
- + Sophisticated interpretation of the flag icon integrated into the text.
- − Failed to render the word 'SUSHI' entirely.
- − Text repetition occurs with 'JAPAN' and 'JAPAN' (partially obscured by circle).
Stable Diffusion 3.5 Large Turbo
- + Successfully rendered 'JAPAN' in a bold, stylized format.
- + Rice texture is more detailed and realistic for a miniature model.
- + Good adherence to the 'raised diorama base' instruction with a wooden block.
- − Spelling error in 'SUSHI' which rendered as 'SIIHI'.
- − The flag icon is generic and resembles the Polish flag rather than the Japanese flag.
- − Background is a gradient dark blue rather than the requested 'solid light blue'.
Verdict: FLUX.1 [schnell] FP8 followed the aesthetic and background requirements much better, creating a clean, professional-looking isometric render, though it failed to include all the requested text. Stable Diffusion 3.5 Large Turbo attempted the text and diorama elements more literally but suffered from spelling errors, an incorrect flag, and a background color that ignored the prompt. FLUX.1 is the winner for its superior visual quality and adherence to the isometric style.
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Includes all four requested animals (dog, cat, bunny, and fox) although species are somewhat blended.
- + Captures the warm golden sunrise light and 'god rays' mentioned in the prompt effectively.
- + High level of fur detail and many butterflies create a rich, complex scene.
- − Biological inaccuracies where some animals look like hybrids (e.g., the fox/cat blend in the bottom right).
- − The anatomy of the animal in the center and bottom right is slightly confused/mangled.
Stable Diffusion 3.5 Large Turbo
- + Clean, vibrant colors with high contrast and sharp focus on the subjects.
- + Very expressive, large eyes that match the 'expressive' requirement of the prompt.
- + Good backlighting effect on the fur edges.
- − Failed to include all four requested animals, missing the baby bunny and red fox kit.
- − The fur textures look more like digital painting/smoothing rather than 'hyper-photorealistic' hair.
- − The composition is much simpler and ignores several elements of the prompt like 'tumbling together'.
Verdict: FLUX.1 [schnell] FP8 is the clear winner as it attempted the full complexity of the prompt, including all four animal types and the specific lighting conditions like god rays. Stable Diffusion 3.5 Large Turbo failed on prompt adherence by omitting half of the requested animals and produced an image that looks more like a 3D render than a photorealistic scene.
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Captures the minimalist vector emblem style well.
- + Balanced composition with delicate decorative elements.
- + Accurate placement of the 'Est. 1720' text.
- − Significant spelling errors in the primary brand name.
- − The cloche depiction lacks the requested steam effect.
Stable Diffusion 3.5 Large Turbo
- + Includes visible steam rising from the cloche knob.
- + Better spelling accuracy for the restaurant name.
- + Stronger vintage texture application on the background.
- − The cloche looks more like a coffee cup hybrid than a pure dome.
- − Text rendering on 'Florin' contains slight artifacts.
Verdict: Stable Diffusion 3.5 Large Turbo is preferred as it closer aligns with the prompt's spelling and specific visual elements like the steam. While FLUX.1 [schnell] FP8 offers a cleaner vector aesthetic, its failure to spell the name correctly and the omission of the steam effect makes it less effective.
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.”
FLUX.1 [schnell] FP8
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
FLUX.1 [schnell] FP8
- + Strongly adheres to the requested professional flat-vector aesthetic.
- + Includes a clear, sequential timeline that follows the requested steps.
- + Follows the NASA-inspired color palette accurately.
- − Text rendering is largely illegible gibberish after the main title.
- − The 'Saturn V' icon is just a circle, failing to represent the rocket.
- − The icons become repetitive and lack variety.
Stable Diffusion 3.5 Large Turbo
- + Features a more dynamic and artistic composition with attractive illustrations.
- + High visual appeal with a retro-modern feel.
- + Good use of space and depth.
- − Fails to follow the specific sequential steps requested in the prompt.
- − Contains odd graphical artifacts like an astronaut standing on a ladder in open space.
- − Misspells 'Apollo' as 'Apoll.o' and uses nonsensical icons for the labels.
Verdict: FLUX.1 [schnell] FP8 is the clear winner for prompt adherence, as it correctly structures the infographic as a six-step journey with a clean, vector look. While Stable Diffusion 3.5 Large Turbo produces a more visually striking illustration, it fails to follow the logical sequence of the mission steps and includes several illogical compositional elements.
Next steps
Explore each model
Distilled version of SD 3.5 Large that generates high-quality images in just 4 steps, offering faster inference and reduced costs