Black Forest Labs' compact, open-source image generation model with sub-second inference, optimized for production and near real-time applications with multi-reference support
Settled by community votes across 7 shared challenges, with an AI judge weighing in on each.
FLUX.2 [klein] 4B
#32 of 62 in Text-to-Image
Qwen Image Edit 2509
#32 of 32 in Image Editing
Where the votes landed
FLUX.2 [klein] 4B
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 [klein] 4B
- + Excellent preservation of the man's facial features and specific hairstyle from the source image.
- + Maintains the plaid pattern of the man's coat on his arm inside the car.
- + High fidelity in preserving the specific car model and details from the source.
- − The transition between the man's arm and the car door is slightly blurry.
Qwen Image Edit 2509
- + Dynamic composition with a strong sense of speed and beautiful lighting.
- + Accurate coastline background that fits the 'California' description well.
- − Fails to preserve the specific facial features and clothing of the person from the source image.
- − The man's hair appears motion-blurred but lacks the distinctive dreadlock structure of the original.
Verdict: FLUX.2 [klein] 4B is the clear winner as it successfully merges both source images, preserving the specific identity of the man (face, hair, coat) and the car. Qwen Image Edit 2509 produces an aesthetically pleasing image but fails the core editing task by replacing the man with a generic figure and losing the source preservation.
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 [klein] 4B
- + Successfully transferred the character's face, sunglasses, and scarf into the pose.
- + Maintained the lighting and background from the source image.
- + Correctly changed the clothing color from red to black to match the character reference.
- − Failed to preserve the hair length/style from Image 2, instead merging it with the hair from Image 1.
- − The hand anatomy on the right arm is distorted with an extra thumb-like digit.
Qwen Image Edit 2509
- + Perfectly preserved the background and pose structure of Image 1.
- − Failed entirely to incorporate the character from Image 2.
- − Only applied a minor filter/shading pass to the original Image 1 without any content edits.
- − The skin tones and textures became overly smoothed and loses natural detail.
Verdict: FLUX.2 [klein] 4B followed the complex instruction of combining a specific character with a specific pose, successfully mapping the face and clothing details onto the new body. Qwen Image Edit 2509 failed to perform the edit requested, returning a filtered version of the first source image without incorporating any elements from the second.
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 [klein] 4B
- + Excellent preservation of the person's unique skin features and facial identity from Image 1.
- + High-fidelity recreation of the pea coat and plaid scarf textures.
- + Natural integration of lighting and shadows on the clothing and skin.
- − Missed several requested accessories including the sunglasses.
- − The scarf pattern is slightly modified compared to the source image.
Qwen Image Edit 2509
- + Successfully included the sunglasses from Image 2.
- + Captures the full-body outfit ensemble including the shoes.
- + Maintains the background environment well.
- − Partially altered the person's facial features and added facial hair not present in Image 1.
- − Lower visual quality with visible artifacts and 'mushy' textures on the face and clothes.
- − Poor placement of the sunglasses which don't align correctly with the bridge of the nose.
Verdict: FLUX.2 [klein] 4B is the clear winner for its superior ability to preserve the subject's identity and skin details from Image 1 while realistically rendering the new clothing. Although Qwen Image Edit 2509 attempted to include more accessories like the sunglasses and shoes, it failed to keep the subject's face unchanged and suffered from significant loss in image clarity and realistic texture.
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 [klein] 4B
- + Excellent source preservation of the face, glasses, and jacket.
- + The hairline and hair texture look very natural and well-integrated.
- + The flyaway hairs add to the realism of the outdoor setting.
- − None notable; the edit perfectly matches the prompt and original style.
Qwen Image Edit 2509
- + Successfully added thick hair while maintaining facial features.
- + Good preservation of the original jacket and background.
- − The hair has a slightly 'plastic' or overly groomed texture compared to the original photo's grit.
- − The hairline looks a bit less natural than model A.
Verdict: Both models did an excellent job of preserving the source image while adding hair. FLUX.2 [klein] 4B is the winner because the hair texture and integration feel much more realistic and match the rugged lighting of the original photo, whereas Qwen Image Edit 2509 produced hair that looks slightly artificial.
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 [klein] 4B
- + Excellent caricature style with professional-looking line art and coloring.
- + Strong depiction of the TV anchor profession with a detailed news desk setting.
- + The dogs are well-integrated into the scene and match the art style perfectly.
- − Completely missed the 'hockey' requirement.
- − Changed the subject's outfit from a denim shirt to a formal suit.
Qwen Image Edit 2509
- + Successfully incorporated all prompt elements: TV anchor, dog, and hockey.
- + Preserved the denim shirt and 'selfie' pose from the original image.
- + Playful composition that blends a news desk with a hockey rink background.
- − The dog is very small and lacks the detail/character seen in the other model.
- − The background hockey players have slightly strange facial features.
- − Text in the speech bubble is nonsensical.
Verdict: Qwen Image Edit 2509 is the winner because it successfully incorporated all three thematic elements requested (TV anchor, dogs, and hockey), whereas FLUX.2 [klein] 4B completely omitted the hockey requirement. Qwen Image Edit 2509 also did a better job of preserving the source image context by keeping the subject's denim shirt and her specific selfie-arm pose.
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 [klein] 4B
- + Excellent adherence to the pastel, watercolor texture request.
- + Creates a very dreamy and nostalgic atmosphere consistent with Studio Ghibli backgrounds.
- + Maintains the composition while softening the features effectively.
- − The character on the right loses the angry expression, appearing too pleasant compared to the source.
- − Slightly too washed out, losing some definition in the forms.
Qwen Image Edit 2509
- + Successfully preserves the specific facial expressions from the meme, particularly the girlfriend's scowl.
- + Good balance between the illustrative style and the original color palette.
- + Keeps the background architectural details more recognizable.
- − The texture feels more like a digital screen tone or halftone dot than a Ghibli-style hand-painted texture.
- − Lighting is less 'dreamy' compared to Model A.
Verdict: Both models successfully translated the 'Distracted Boyfriend' meme into an illustrative style. FLUX.2 [klein] 4B followed the stylistic 'Ghibli' prompt more accurately with its soft watercolor textures and dreamy lighting, though it failed to preserve the core emotion of the jealous girlfriend. Qwen Image Edit 2509 preserved the original scene's humor and expressions much better, but its texture feels more like a generic manga pattern than a hand-painted Ghibli background.
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 [klein] 4B
- + Excellent source preservation, keeping the person and dog identical to the original.
- + The leaves look realistic and vary in depth and motion blur.
- + The hair edit is subtle and follows natural physics.
- − The wind effect on the hair is slightly less 'energetic' than the prompt might suggest.
Qwen Image Edit 2509
- + Successfully added a very dynamic wind effect to the hair.
- + Added colorful leaves as requested.
- − The leaves appear like 2D stickers and do not match the lighting or depth of the scene.
- − The hair rendering looks slightly artificial and stringy compared to the source.
- − The dog's tail has been modified unnecessarily, losing its original shape.
Verdict: FLUX.2 [klein] 4B is the clear winner because it maintains the integrity of the source image perfectly while adding realistic, depth-aware leaves and a natural hair-blowing effect. Qwen Image Edit 2509 creates a more stylized hair effect but fails on the implementation of the leaves, which appear pasted on and lack environmental integration.
Explore each model
Alibaba's Qwen image editing model for instruction-based image modifications and transformations