Fast distilled version of Black Forest Labs' FLUX.2 [dev] optimized for speed and cost efficiency.
Settled by community votes across 7 shared challenges, with an AI judge weighing in on each.
FLUX.2 [dev] Flash
#7 of 62 in Text-to-Image
Qwen Image Edit 2509
#32 of 32 in Image Editing
Where the votes landed
FLUX.2 [dev] Flash
0%
win rate
Ties
0%
Qwen Image Edit 2509
0%
win rate
Challenge by challenge
The strongest take from each model on every shared challenge, with the AI judge's read.
Man and Car in California
Editing“Make a photo of the man driving the car down the California coastline”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Excellent preservation of the subject's face, hair, and unique clothing style.
- + Great attention to detail on the car's interior wooden dash and leather seats.
- + Realistic perspective captured from inside the vehicle.
- − The car interior, while high quality, differs slightly from the convertible shown in the source.
- − The steering wheel is a generic vintage style rather than the Rolls-Royce style.
Qwen Image Edit 2509
- + Perfect preservation of the specific Rolls-Royce car model from the source image.
- + Captures a wide, scenic view of the California coastline as requested.
- − The man's face and distinctive clothing from the source image are largely lost or changed.
- − Poor person-to-car scale; the driver appears much too small for the vehicle.
Verdict: FLUX.2 [dev] Flash is the clear winner because it successfully merges both source elements: it places the specific person from the second source image into a car on the coastline while maintaining his unique clothing and facial features. While Qwen Image Edit 2509 preserved the car perfectly, it failed to accurately represent the man and created a jarring scale issue between the driver and the vehicle.
Pose & Character Mashup
Editing“Use Image 1 as the exact pose reference and Image 2 as the character reference. Recreate the person/character from Image 2 in the exact dynamic pose and body position from Image 1. Keep the exact face, hair, clothing style/details, and expression from Image 2. Match the lighting and environment of Image 1. The final image must show the character from Image 2 performing the precise action/pose from Image 1 with perfect anatomy and natural integration.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Successfully integrated the clothing and scarf from Image 2
- + Accurately captured the face and sunglasses of the character from Image 2
- + Matched the yellow studio lighting and background of the original scene
- − Failed to match the exact pose, showing the character upright instead of leaning forward
- − Severe anatomical glitches including a duplicate ghostly face and hair coming out of the back
- − The feet are floating or placed awkwardly on the stool
Qwen Image Edit 2509
- + Maintained the exact pose from Image 1 perfectly
- − Completely failed to incorporate the character reference from Image 2
- − No change to the face, hair, or clothing style from the original person in Image 1
- − Applied an unnecessary artistic filter that reduced image clarity
Verdict: Both models struggled significantly with this complex instruction. FLUX.2 [dev] Flash attempted to merge the character details but failed the pose and produced a nightmare-like anatomical error with multiple heads. Qwen Image Edit 2509 perfectly preserved the pose but ignored the instruction to change the character, effectively failing the core editing task entirely. FLUX.2 is slightly better for at least attempting the character swap, despite the poor execution.
Outfit Transfer Challenge
Editing“Use Image 1 as the base person. Dress them in the exact elaborate outfit from Image 2 (including all layers, accessories, jewelry, and shoes). Carefully adapt the clothing to the body shape and pose in Image 1 while maintaining realistic fabric behavior, correct proportions, and perfect lighting/shadow matching. Keep the person’s exact face, hair, and background completely unchanged.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Successfully transferred the coat and scarf textures.
- + Identified and added elaborate jewelry mentioned in the prompt.
- − Severely distorted the person's face into a double-exposure nightmare.
- − Failed to maintain the person's identity and background integrity.
- − Incorrectly merged the two subjects' faces.
Qwen Image Edit 2509
- + Successfully applied the coat, scarf, and jeans from Image 2 to the person from Image 1.
- + Maintained the overall identity and skin patterns of the person in the base image.
- + Handled the pose adaptation and full-body composition reasonably well.
- − Lower visual quality with some painterly/smudged artifacts.
- − The sunglasses slightly obscure the original person's eyes.
- − Added shoes that were not visible in the source image, though they fit the style.
Verdict: FLUX.2 [dev] Flash failed the primary constraint of keeping the person's face and hair unchanged, instead creating a horrifying composite of both individuals. Qwen Image Edit 2509 successfully completed the edit by placing the wardrobe on the target person while preserving their identity and the background.
Bald man challenge
Image Editing“Give the person a full, thick head of natural hair with realistic texture, density, and a natural hairline. Preserve facial features and lighting.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Excellent preservation of the source image facial features and background
- + Matches the lighting of the scene perfectly
- + Follows the instruction for a thick head of hair with realistic texture
- − The transition at the hairline is slightly harsh
- − The hair texture choice is a bit stylistically mismatched with the straighter beard hair
Qwen Image Edit 2509
- + Natural and realistic hairline integration
- + Hair texture more closely matches the aesthetic of the beard
- + High level of preservation for the rest of the image
- − The hair has some oily or 'plastic' sheen artifacts in the highlights
Verdict: Both models handled the edit exceptionally well, preserving the person's identity and the surrounding environment almost perfectly. Qwen Image Edit 2509 is the winner because the hairline and choice of hair style feel much more integrated and natural for the subject than the afro selected by FLUX.2 [dev] Flash, despite some minor shiny artifacts in the hair strands.
Over-the-top cartoon caricature
Editing“Create a caricature of me and my job. Make it exaggerated and humorous, incorporating my profession as a tv show anchor and my love for dogs and hockey.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Excellent caricature style with great facial likeness preservation
- + Perfectly integrates all requested elements: TV anchor desk, hockey rink, and multiple dogs
- + High visual quality with clean lines and professional-looking illustration
- − The transition from the small body to the large head is a bit generic for high-end caricatures
Qwen Image Edit 2509
- + Maintains the selfie perspective and denim outfit from the source image
- + Incorporates the hockey and TV elements effectively into the background
- − Failed to include the dogs in a meaningful way (only one tiny, toy-like dog)
- − The facial likeness to the source person is significantly distorted/lost
- − Text in the speech bubble and the background screen is gibberish
Verdict: FLUX.2 [dev] Flash significantly outperformed the competitor by successfully integrating all three requested themes (TV anchor, hockey, and dogs) into a cohesive and high-quality caricature. While Qwen Image Edit 2509 preserved the specific clothing and pose of the original photo, it failed on the likeness preservation and included very poor quality text and minimal dog representation.
Studio Ghibli Anime Style
Editing“Transform this photo into a Studio Ghibli–inspired illustration. Use soft pastel colors, hand-painted textures, gentle lighting, dreamy backgrounds, and a warm, nostalgic mood”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Expertly captures the hand-painted Ghibli texture and lighting.
- + Preserves the exact poses and facial expressions of the source image while stylizing them.
- + The dreamy meadow background adds to the nostalgic atmosphere requested.
- − Changes the background setting from a city street to a field, which may deviate from strict source preservation.
Qwen Image Edit 2509
- + Retains the original city background elements like the bus and street layout.
- + Good color palette that matches the 'soft pastel' requirement.
- + Preserves the composition and poses very accurately.
- − The art style leans more toward modern webtoon or digital vector art than Ghibli's hand-painted aesthetic.
- − Background execution looks a bit unfinished or flat compared to the foreground.
Verdict: FLUX.2 [dev] Flash and Qwen Image Edit 2509 both did an excellent job of maintaining the iconic 'distracted boyfriend' pose and composition. FLUX.2 [dev] Flash is the clear winner for the specific Ghibli prompt, as it perfectly mimics the watercolor texture and soft, blooming light characteristic of the studio, whereas Qwen Image Edit 2509 feels like a generic digital illustration.
Golden Hour Stroll
Image Editing“Add dynamic motion to this photo: make hair blow in the wind, add leaves flying, energetic and lively feel.”
AI Judge Analysis
FLUX.2 [dev] Flash
- + Excellent preservation of the subject's face and jacket details.
- + Natural-looking hair flow that blends well with the original hairline.
- + Large quantity of leaves that feel integrated into the 3D space of the scene.
- − The leaf motion is a bit static, lacking motion blur to sell the 'dynamic' request.
- − One leaf is awkwardly resting on the subject's chest.
Qwen Image Edit 2509
- + Creative addition of motion to the dog's tail via a 'fanned' effect.
- + Hair has a more distinct wind-swept shape.
- − The added leaves look like 2D stickers and do not match the lighting or depth of the scene.
- − The subject's face has been slightly altered/distorted compared to the source image.
- − Strange artifacts in the blowing hair texture that look like 'ribbons'.
Verdict: FLUX.2 [dev] Flash is the winner as it successfully adds the requested elements while maintaining the high quality and identity of the original image. While Qwen Image Edit 2509 attempted more movement (like the dog's tail), its leaf additions look very artificial and it failed to preserve the subject's facial likeness as well as the Flux model.
Explore each model
Alibaba's Qwen image editing model for instruction-based image modifications and transformations