Distilled version of Black Forest Labs' FLUX.2 [dev] outperforming it at a cheaper price. Developed by fal.ai.
Settled by community votes across 7 shared challenges, with an AI judge weighing in on each.
FLUX.2 [dev] Turbo
#3 of 62 in Text-to-Image
Qwen Image Edit 2509
#32 of 32 in Image Editing
Where the votes landed
FLUX.2 [dev] Turbo
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] Turbo
- + Excellent preservation of the subject's face, hair, and clothing.
- + The composition provides an intimate, realistic view of the car interior.
- + Maintains logical lighting between the subject and the new environment.
- − The car color has shifted from white to a more silver/beige tone.
- − The scale of the subject relative to the car seems slightly large.
Qwen Image Edit 2509
- + Perfectly preserves the white color and exterior details of the car.
- + Captures a wide, scenic view of the California coastline.
- + Good sense of motion with the blurred road.
- − The man is barely visible and the likeness is poorly preserved compared to the source.
- − The man's hair is blowing forward in a way that defies physics if the car is moving forward.
Verdict: FLUX.2 [dev] Turbo succeeded much better in the primary task of placing the specific man from the source image into the car, maintaining his clothing and facial features with high fidelity. While Qwen Image Edit 2509 preserved the car's color more accurately, the subject became a generic figure in the distance, failing the character preservation aspect of the edit.
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] Turbo
- + Successfully transferred the character's clothing, scarf, and face from Image 2.
- + Maintained the yellow background and red ottoman from Image 1.
- + High resolution and clear detail on the character's features.
- − Anatomical failure with a second head growing from the character's side.
- − Did not accurately replicate the specific pose from Image 1, creating a generic floating pose instead.
Qwen Image Edit 2509
- + Successfully preserved the original Image 1 with no significant changes to the scene.
- − Failed the core task by completely ignoring the character reference from Image 2.
- − Simply applied a slight filter/re-render to Image 1 without changing the person.
Verdict: FLUX.2 [dev] Turbo attempted the complex task of character and outfit transfer, successfully bringing over the man's face and scarf, though it suffered a major anatomical glitch with a second head. Qwen Image Edit 2509 failed the prompt entirely by returning Image 1 with no character changes.
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] Turbo
- + Successfully captured the exact styling of the jacket, scarf, and sunglasses from Image 2.
- + High image resolution and clarity in textures.
- + Perfectly matched lighting on the person to the background environment.
- − Completely failed to use the person from Image 1, essentially pasting the person from Image 2 onto the background.
- − Violated the instruction to keep the person's exact face and hair unchanged.
- − Added excessive, unrequested jewelry that was not in either source image.
Qwen Image Edit 2509
- + Successfuly used the person from Image 1 as the base, preserving the skin condition and general facial structure.
- + Maintained the background from Image 1 and applied the clothing from Image 2 realistically to the pose.
- + Attempted to include shoes as requested by the prompt.
- − Overall image quality is noticeably lower with some blurring and artifacts.
- − The face has been slightly altered/morphed beyond the original person in Image 1.
- − Failed to preserve the exact hair from Image 1, making it shorter and darker.
Verdict: This is a case of two different failures. FLUX.2 [dev] Turbo produced a higher-quality image but completely ignored the primary instruction to use the person from Image 1, simply moving the person from Image 2 to a new background. Qwen Image Edit 2509 followed the core identity instruction much better by keeping the person with vitiligo, but suffered from lower fidelity and some facial morphing; however, it remains the better edit as it actually combined the subjects correctly.
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] Turbo
- + Excellent preservation of the source image face and lighting.
- + Hair texture realistically matches the light and environment.
- − The hairline transition on the forehead is a bit harsh.
- − The hairstyle chosen is quite dramatic compared to the original aesthetic.
Qwen Image Edit 2509
- + Natural-looking wavy hair texture.
- + Great integration of the hair with the existing sideburns and beard.
- − The hair appears slightly 'pasted on' at the very top with slightly different lighting than the scalp.
- − Minor alteration of the forehead shape.
Verdict: Both models did an excellent job of preserving the original image's character, lighting, and background. FLUX.2 [dev] Turbo provided a very high-density afro-style edit that feels physically grounded, while Qwen Image Edit 2509 provided a shorter, wavy style that feels slightly more integrated with the subject's existing beard color and texture.
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] Turbo
- + Excellent caricature style with professional-level shading and detail.
- + Perfectly hits all prompt requirements: news anchor desk, hockey rink background, and multiple dogs.
- + Maintains strong facial resemblance to the subject in a stylized way.
- − The text on the anchor desk is gibberish.
- − The subject is wearing a suit instead of the denim jacket from the source image.
Qwen Image Edit 2509
- + Successfully preserves the blue denim jacket from the source photo.
- + Good use of the 'selfie' arm perspective from the original image.
- + Clear caricature style that captures the subject's expression.
- − The dog is very small and lacks detail compared to the rest of the image.
- − The gibberish text in the speech bubble and on monitors is distracting.
- − General image quality is slightly flat with less sophisticated rendering than Model A.
Verdict: FLUX.2 [dev] Turbo produced a much more comprehensive and high-quality illustration that better integrated the hockey and dog elements into a cohesive scene. While Qwen Image Edit 2509 did a better job of preserving the source clothing (the denim jacket), its execution of the dogs was weak and the overall composition felt less professional.
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] Turbo
- + Excellently captures the watercolor, hand-painted texture synonymous with Studio Ghibli.
- + Maintains character poses and composition with high fidelity.
- + Dreamy background with cherry blossoms fits the requested aesthetic perfectly.
- − The woman in the foreground is slightly more centered than in the original crop.
Qwen Image Edit 2509
- + Successfully translates the image into a clean anime style.
- + Preserves the spatial arrangement and character expressions accurately.
- − Lacks the requested hand-painted texture, appearing more like modern digital cel-shading.
- − The halftone dot pattern in the background and skin is not characteristic of Studio Ghibli.
- − The color of the dress is significantly faded compared to the original.
Verdict: FLUX.2 [dev] Turbo is the clear winner for this specific task as it successfully incorporated the 'hand-painted textures' and 'dreamy' Ghibli aesthetic while maintaining the structure of the iconic meme. Qwen Image Edit 2509 produced a clean animation style but relied on digital artifacts like halftone patterns that contradicted the request for a nostalgic, painted mood.
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] Turbo
- + Excellent preservation of the subject's face and original details
- + Natural integration of wind-blown hair that matches the lighting
- + Consistent leaf style that looks realistic within the scene
- − The leaf motion blur is consistent but some leaves overlap the subject's body awkwardly
Qwen Image Edit 2509
- + Successfully added motion to the dog's tail in addition to the woman's hair
- + Vibrant, high-contrast leaves add to the energetic feel
- − The hair effect looks slightly more artificial/digital than the source
- − Noticeable changes to the subject's facial features and eyes compared to the original
Verdict: FLUX.2 [dev] Turbo followed the editing instructions while maintaining near-perfect fidelity to the original person's face and the dog's appearance. Qwen Image Edit 2509 added a nice touch by animating the dog's tail, but it struggled with source preservation, noticeably altering the woman's facial features and eye shape.
Explore each model
Alibaba's Qwen image editing model for instruction-based image modifications and transformations