GPT Image 2 OpenAI vs Qwen Image Alibaba

GPT Image 2

• + Excellent photographic texture on the book cover and wooden table.
• + Highly realistic refraction and transparency in the glass cube.
• + Accurate handling of the plant's visibility through the glass.

• − The glass cube has internal structural seams that make it look a bit like a frame rather than a solid glass object.

Qwen Image

• + Successfully includes all requested elements in a clean composition.
• + Good reflection of the blue sphere on the bottom surface of the cube.
• + Soft window lighting is accurately portrayed.

• − Lower overall resolution and sharpness compared to Image A.
• − The plant's appearance through the glass is slightly inconsistent with the leaves behind it.
• − The wooden texture on the table is less detailed.

GPT Image 2

• + Excellent skin texture and hyper-realistic facial details
• + Accurate representation of 1990s-style Japanese city streets and signage
• + Natural 'imperfect' framing that feels like a real candid photo

• − The rain effect is very subtle, almost difficult to see
• − The motion blur on the car in the background is slightly inconsistent with the depth of field

Qwen Image

• + Stronger visual representation of 'light rain' and atmospheric mood
• + Good use of reflections on the wet pavement
• + Solid composition with clear depth of field

• − The facial features and hair look slightly artificial and smoothed out
• − The bicycle has structural issues, such as a kickstand floating in mid-air
• − Lower image resolution and loss of fine texture compared to Model A

GPT Image 2

• + Exceptional photorealistic skin texture and eye detail
• + Incredibly intricate engraving and weathered patina on the armor
• + Natural-looking braids with subtle, believable bead integration

• − The 'bokeh sparks' are extremely subtle, almost unnoticeable
• − Background is slightly generic

Qwen Image

• + Strong adherence to the 'beads' and 'bokeh sparks' prompt elements
• + Good depiction of scarred skin and varied textures
• + High contrast lighting captures the torch effect well

• − The sparks look like artificial icons rather than natural lens bokeh
• − Skin and hair textures are slightly smoothed and 'CG' in appearance
• − Beads look more like modern plastic carnival beads than historical/fantasy accessories

GPT Image 2

• + Exceptional text rendering with coherent item names, descriptions, and prices.
• + Highly professional layout that fulfills all prompt requirements for specific sections.
• + High-quality, distinct food photography for every menu item.

• − The layout is very dense, which may slightly push the boundaries of 'minimalist' compared to a simpler grid.

Qwen Image

• + Strong minimalist aesthetic with a clear grid-based design.
• + Interesting use of vibrant background colors behind food photography.

• − Text is largely gibberish and poorly rendered.
• − Failed to create three distinct sections, merging pizza and mains into one header.
• − Lower resolution food images with repeating visual elements.

GPT Image 2

• + Excellent adherence to the fiery/glowing text prompt across all three elements.
• + High-quality textures and realistic lighting on food components.
• + Strong dynamic composition with a clear 'exploded' view.

• − The text takes up a very large portion of the frame compared to the product.

Qwen Image

• + Clean, readable text layout that feels more like a commercial poster.
• + Bright, appetizing colors on the burger components.

• − Failed to render the 'LIMITED TIME ONLY' and '€6.99' text with the requested fiery, glowing effect.
• − The burger is less 'exploded' and more of a whole burger with a few extra pieces floating around it.
• − The price tag lacks the fiery glow requested in the prompt.

GPT Image 2

• + Excellent chalk texture with realistic dusty edges and pressure variations.
• + Perfect adherence to the requested text and date.
• + Highly realistic cursive handwriting that truly looks manually written.

• − The lighting in the upper left corner slightly washes out a small portion of the text.

Qwen Image

• + Clear, legible text layout.
• + Good use of space on the board.

• − Failed the date requirement by writing '20026' instead of '2026'.
• − The text looks like a smooth digital font with a slight glow rather than realistic chalk.
• − Text lacks the authentic chalk texture and physical grit requested in the prompt.

GPT Image 2

• + Excellent adherence to the 'horse on top' spatial instruction
• + High textural detail in the spacesuit and lunar surface
• + Clever integration of riding tack including stirrups for the horse's hooves

• − The astronaut's hands/gloves have a somewhat unnatural, finger-heavy appearance

Qwen Image

• + Good cinematic lighting and planet design
• + Clean rendering of the astronaut and horse components

• − Completely failed the negative constraint to have the horse on top
• − Anatomical issues with the horse's legs appearing disjointed

GPT Image 2

• + Excellent fur texture and lighting integration for the capybara
• + High photographic realism in the interior and background bokeh
• + Accurate depiction of a professional 'NY taxi driver' expression

• − The passenger's hands and phone interaction are slightly blurry and less defined

Qwen Image

• + Clear composition showing the full scene including the taxi's exterior and interior
• + Good adherence to the 'bored' expression requested for the passenger
• + Includes the exterior taxi roof light for additional context

• − The capybara's hands look more like monkey hands/paws than capybara paws
• − The capybara's head is fused strangely with the car seat headrest
• − Internal lighting is a bit flat compared to the realism of Model A

GPT Image 2

• + Excellent typography with perfect spelling for all requested text.
• + Highly detailed and cohesive vintage aesthetic with intricate borders and cinematic lighting.
• + Smart inclusion of a bridge background to represent 'The Arches' location.

• − The parchment background is very dark, which might make some side details harder to see.

Qwen Image

• + Strong contrast between the parchment edge and the dark central scene.
• + Follows the general composition requested, including the pumpkin and twisted trees.

• − Significant spelling errors in the main title ('Halle Party Invitation') and extra gibberish text at the top.
• − Lower level of artistic detail compared to Model A.
• − The 'scroll' banner is broken into two pieces, reducing the polished look.

GPT Image 2

• + Excellent PBR textures and lighting that create a high-end 3D render look.
• + Perfect text rendering and placement as requested.
• + Highly detailed and realistic sushi models within the diorama context.

• − The diorama content is quite crowded, pushing the definition of 'minimal garnish'.

Qwen Image

• + Successfully captures the requested 'cartoon miniature' style with a soft, clean aesthetic.
• + Good adherence to the diorama base and minimal garnish instructions.
• + Correct 45° isometric perspective.

• − Text rendering is slightly messy with overlapping and uneven spacing.
• − The small flag icon in the text section is distorted and lacks the clean execution of Model A.

GPT Image 2

• + Excellent depiction of dynamic movement with realistic paws and poses
• + Very realistic fur texture and lighting integration on all animals
• + Includes all requested animals with distinct, accurate features

• − The fox's limb positioning looks a bit awkward in the mid-leap
• − The background butterflies are a bit blurry compared to the foreground

Qwen Image

• + Beautiful dew sparkles and soft bokeh effect in the grass
• + Cute, expressive eyes on all animals
• + Great realization of the sunset god rays

• − The fox has a cat-like paw structure and slightly stylized features
• − The golden retriever puppy looks static compared to the others
• − Some minor blending artifacts where the rabbit meets the dog

GPT Image 2

• + Excellent typography with perfect spelling of 'Caffè Florian'.
• + Sophisticated vintage engraving style with high-quality cross-hatching.
• + Balanced composition with a professional emblem frame and banner.

• − Slightly more ornate than 'minimalist', though it fits the 'vintage' prompt perfectly.

Qwen Image

• + Stronger adherence to the 'minimalist' aspect of the prompt.
• + Correct color palette of warm browns and cream.

• − Severe typography errors with cluttered, overlapping text ('FLoraIAN').
• − The graphic style is a bit too simplified and lacks the requested 'classic' feel.

GPT Image 2

• + Excellent typography and legible text rendering throughout the infographic.
• + Highly detailed and accurate visual representations of the Saturn V and Lunar Module.
• + Comprehensive layout that follows all six requested steps in a logical sequence.

• − The illustration style leans more toward 3D rendering than the 'flat-vector' style requested.
• − Slightly more complex gradients than the 'subtle' ones requested.

Qwen Image

• + Accurately captures the requested flat-vector aesthetic with simple shapes.
• + Adheres strictly to the NASA-inspired color palette requested in the prompt.

• − Severe spelling and text rendering errors ('ApolL', 'Sarth Orbit', 'Arnnstrong').
• − The composition is cluttered and fails to logically follow the requested six-step sequence.
• − Included meta-instructions in the final image text like '(Stop at landing)'.