Black Forest Labs' precision image generation model with maximum control, reliable text rendering, and complete creative control supporting up to 4MP output
Settled by community votes across 6 shared challenges, with an AI judge weighing in on each.
FLUX.2 [flex]
24.8
arena score
#14 of 62 in Text-to-Image
Skill signature
· Text-to-Image
Qwen Image Edit 2511
20.2
arena score
#16 of 32 in Image Editing
Vote tally
Where the votes landed
FLUX.2 [flex]
49.2%
win rate
Ties
0.0%
Qwen Image Edit 2511
50.8%
win rate
49.2%
0.0% ties
50.8%
Shared challenges
6
Challenge by challenge
The strongest take from each model on every shared challenge, with the AI judge's read.
Man and Car in California
EditingEdit instruction
“Make a photo of the man driving the car down the California coastline”
Source
FLUX.2 [flex]
Qwen Image Edit 2511
49% wins
0% ties
51% wins
AI Judge Analysis
FLUX.2 [flex]
- + Excellent preservation of the specific Rolls-Royce model from the source image
- + Natural integration of the man into the driver's seat
- + High-quality motion blur on the wheels and road suggests realistic driving
- − The man's clothing is not perfectly preserved from the source image
- − The man's facial features are slightly simplified due to the distance
Qwen Image Edit 2511
- + Near-perfect preservation of the man's face, hair, and clothing (plaid coat and scarf)
- + Excellent composition showing both the interior details of the car and the scenic background
- + Accurate depiction of a California coastal highway with palm trees
- − The car interior/dashboard does not perfectly match the specific Rolls-Royce model from the source
- − The steering wheel placement looks slightly cramped relative to the driver
Verdict: Qwen Image Edit 2511 is the winner because it successfully preserves the specific identity and attire of the man from the source image, whereas FLUX.2 [flex] modifies his clothing. While FLUX.2 [flex] did a better job of keeping the exterior shell of the specific car model, Qwen Image Edit 2511's ability to maintain the subject's appearance and the requested setting in a high-quality composition is more impressive for a complex image-to-image edit.
Pose & Character Mashup
EditingEdit instruction
“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.”
Source
FLUX.2 [flex]
Qwen Image Edit 2511
AI Judge Analysis
FLUX.2 [flex]
- + Successfully integrated the character, clothing, and accessories from Image 2 into a yellow studio setting.
- + Captured the yellow background and red ottoman from Image 1 correctly.
- + Followed the instruction to apply the character's face, hair, and scarf to the new scene.
- − The pose is a generic crouching squat rather than the specific dynamic leg-crossing pose from Image 1.
- − The anatomy of the right hand and feet is somewhat distorted.
Qwen Image Edit 2511
- + Retained both original subjects in a composite format.
- − Completely failed to perform the pose transfer, instead just layering the two images together.
- − Ignored the instruction to match the lighting and environment of Image 1.
- − The execution is a low-quality collage with visible seams and artifacts.
Verdict: FLUX.2 [flex] successfully attempted the multi-step instruction by placing the character from Image 2 into the environment and lighting of Image 1, though it simplified the complex pose. Qwen Image Edit 2511 failed the task entirely, producing a crude collage that did not transfer the character into the pose or the new environment.
Bald man challenge
Image EditingEdit instruction
“Give the person a full, thick head of natural hair with realistic texture, density, and a natural hairline. Preserve facial features and lighting.”
FLUX.2 [flex]
Qwen Image Edit 2511
AI Judge Analysis
FLUX.2 [flex]
- + Excellent preservation of the original facial features and bone structure.
- + Perfectly matches the lighting and color tone of the beard with the new hair.
- − The hair length is quite short, bordering on a buzz cut rather than a 'full, thick' head of hair.
- − The top edge of the hair feels slightly flat and lacks some depth.
Qwen Image Edit 2511
- + Successfully added a very full, thick head of wavy hair as requested.
- + Realistic hair texture and high density.
- − Significantly altered the shape of the face, making it look wider and changing the forehead region.
- − The blend between the new hair and the existing glasses frame is messy, especially on the right side of the image.
Verdict: FLUX.2 [flex] produced a much more convincing edit by keeping the original person's identity perfectly intact and matching the lighting, though the haircut is shorter than expected. Qwen Image Edit 2511 followed the prompt for 'thick' hair more literally but failed to preserve the source image's facial structure and introduced artifacting around the glasses.
Over-the-top cartoon caricature
EditingEdit instruction
“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.”
Source
FLUX.2 [flex]
Qwen Image Edit 2511
AI Judge Analysis
FLUX.2 [flex]
- + Expertly incorporates all three thematic elements: TV anchor desk, multiple dogs, and hockey sticks/jerseys.
- + Maintains excellent likeness to the source person within a stylized caricature format.
- + High level of detail in the background, including a news monitor with related imagery and consistent clothing choice.
- − The 'NEWS' text is slightly garbled ('THEIANAES').
- − Overall composition is very crowded.
Qwen Image Edit 2511
- + Successfully captures a vibrant caricature style with exaggerated facial features that still resemble the subject.
- + Clearly depicts the TV anchor profession with a microphone and studio setting.
- − Completely fails to include the hockey theme requested in the prompt.
- − The text in the speech bubble is nonsensical gibberish.
- − Contains strange artifacts like a candle burning on a computer stand.
Verdict: FLUX.2 [flex] is the clear winner as it followed every part of the complex prompt, including the specific request for hockey and dogs alongside the TV anchor profession. While Qwen Image Edit 2511 produced a good caricature stylization, it ignored the hockey requirement and included several nonsensical elements like unreadable text and a random candle.
Studio Ghibli Anime Style
EditingEdit instruction
“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”
Source
FLUX.2 [flex]
Qwen Image Edit 2511
AI Judge Analysis
FLUX.2 [flex]
- + Excellently captures the authentic Ghibli art style with flat line art and watercolor textures
- + Performs a complete stylistic overhaul of the characters while maintaining their poses
- + Perfectly executes the soft pastel color palette and warm, nostalgic mood requested
- − Changes the 'shocked/angry' expression of the girl on the right to a more neutral/smiling one, losing the meme context
Qwen Image Edit 2511
- + Preserves the character likenesses and original facial expressions very accurately
- + Successfully transforms the background into an illustrative style
- + Details like the sunglasses on the red dress are preserved from the original context
- − The art style is more generic 'manhwa' or digital illustration rather than the specific Studio Ghibli style requested
- − The lighting and textures feel a bit too digital and polished to be 'hand-painted'
Verdict: FLUX.2 [flex] provides a far superior transformation into the requested Studio Ghibli style, featuring the characteristic thin line work and watercolor textures associated with the studio, even if it loses some of the specific facial expressions from the source. Qwen Image Edit 2511 maintains the source image structure and expressions better, but its artistic interpretation is a generic digital painting style that fails to capture the specific aesthetic of Studio Ghibli.
Golden Hour Stroll
Image EditingEdit instruction
“Add dynamic motion to this photo: make hair blow in the wind, add leaves flying, energetic and lively feel.”
FLUX.2 [flex]
Qwen Image Edit 2511
AI Judge Analysis
FLUX.2 [flex]
- + Excellent preservation of the original image's identity and background.
- + Realistic hair movement that feels integrated with the head.
- + Effective inclusion of motion-blurred leaves.
- − One of the dog's legs becomes a bit blurry/distorted compared to the source.
Qwen Image Edit 2511
- + Stronger 'energetic' feel with hair spreading out more dramatically.
- + Good variety in leaf colors (orange and green) to suggest a more dynamic scene.
- + High source preservation of the woman and dog's faces.
- − The hair flow looks slightly more artificial and symmetrical.
- − Leaf placement feels a bit flatter against the 2D plane.
Verdict: Both models handled the edit exceptionally well, preserving the original subject matter while perfectly executing the hair and leaf instructions. FLUX.2 [flex] produced a slightly more realistic integration of the blowing hair, whereas Qwen Image Edit 2511 created a more visually striking, energetic effect with a colorful variety of leaves. FLUX.2 [flex] is the winner by a small margin for maintaining a more natural photographic quality in the hair texture.
Next steps
Explore each model
Alibaba's Qwen image editing model for instruction-based image modifications and transformations