OpenAI
976 Elo ±47
Rank #24 of 42 · median 5s per image · 11 refused
Not judged in the same call yet
Elo expectation: ChatGPT 4o wins 49%
Google
980 Elo ±45
Rank #23 of 42 · median 12s per image

By category

CategoryChatGPT 4oImagen 4.0 UltraDirect record
Anime & Cartoon Style 746 999 —
Architecture & Interiors 910 837 —
Complex Scenes 1054 947 —
Ghibli style 813 1111 —
Graphic Design 1099 1092 —
Hands & Anatomy 933 977 —
Photorealistic People & Portraits 1078 922 —
Surreal & Creative Prompts 923 964 —
Text in Images 1003 1008 —
Ultra Hard 1004 978 —

By criterion

CriterionChatGPT 4oImagen 4.0 Ultra
Prompt adherence 1020 1032
Technical quality 964 976
Artistic merit 901 959
Detail execution 924 945
Realism & coherence 995 928

Side by side

A Looney Tunes-inspired cartoon scene of a clever rabbit tricking a hunter. - ChatGPT 4o
ChatGPT 4o 6 / 10 Ranked 9th of 12: this is a well-designed cartoon illustration, but it misses the prompt's central action. Its consistent execution places it above the more awkward images, while its absent trick limits the overall result.
A Looney Tunes-inspired cartoon scene of a clever rabbit tricking a hunter. - Imagen 4.0 Ultra
Imagen 4.0 Ultra 9 / 10 Ranks 2nd of 11, narrowly behind the pitfall image because its comic payoff is still developing. Its strong narrative, expansive composition, and polished drawing make it an excellent result.
A photorealistic interior of a modern Scandinavian-style living room with natural light streaming through floor-to-ceiling windows. Show detailed wood textures, fabric materials, and proper light reflections. - ChatGPT 4o
ChatGPT 4o 8 / 10 Ranked 7th of 13, this is a strong and particularly faithful interpretation of the prompt. It falls behind the leaders mainly in material nuance and visual individuality, rather than through major errors.
A photorealistic interior of a modern Scandinavian-style living room with natural light streaming through floor-to-ceiling windows. Show detailed wood textures, fabric materials, and proper light reflections. - Imagen 4.0 Ultra
Imagen 4.0 Ultra 6 / 10 Rank 8 of 12: this combines strong prompt adherence with several well-rendered materials. The visible lamp defect is more distracting than the small imperfections in the upper-ranked images.
A busy city intersection with cars, pedestrians, and street performers all in one frame. - ChatGPT 4o
ChatGPT 4o 8 / 10 Rank 5 of 13: this is a visually distinctive and strongly adherent result with excellent foreground performers. Its darker presentation and less natural pedestrian details prevent it from matching the best photographic examples.
A busy city intersection with cars, pedestrians, and street performers all in one frame. - Imagen 4.0 Ultra
Imagen 4.0 Ultra 5 / 10 Ranked 11th of 12 because its attractive twilight atmosphere cannot overcome substantial incoherence in the central performance. The prop arrangement is a major weakness in an otherwise well-populated intersection.
A mechanical creature in a polluted landscape. Capture the visual aesthetic of 'Nausicaä' by Studio Ghibli, with Miyazaki's mechanical designs and environmental contrast. - ChatGPT 4o
ChatGPT 4o 6 / 10 Ranks 8th of 11. Its emotional environmental narrative is compelling, but weak structural readability and substantial stylistic deviation prevent a higher placement.
A mechanical creature in a polluted landscape. Capture the visual aesthetic of 'Nausicaä' by Studio Ghibli, with Miyazaki's mechanical designs and environmental contrast. - Imagen 4.0 Ultra
Imagen 4.0 Ultra 8 / 10 Ranks fourth of eleven through strong stylistic fidelity and an unusually integrated mechanical-organic concept. Slightly congested limb construction and tonal overlap prevent it from reaching the leading tier.
Minimalist logo for a sustainable coffee brand 'Evergreen Brew', single leaf motif, clean lines, vector style, white background. 'Evergreen Brew' contained as text in logo - ChatGPT 4o
ChatGPT 4o 6 / 10 Ranked 7th of 13, this is strongly on-brief but comparatively less refined. Heavy typography and inconsistent finishing keep it below the more professionally balanced single-leaf designs.
Minimalist logo for a sustainable coffee brand 'Evergreen Brew', single leaf motif, clean lines, vector style, white background. 'Evergreen Brew' contained as text in logo - Imagen 4.0 Ultra
Imagen 4.0 Ultra 9 / 10 Rank 1 of 12: the strongest combination of literal prompt adherence, clean construction, and balanced typography. Its conventional concept prevents a flawless score, but the execution is near-professional.
A yoga practitioner in a complex pose, arms and legs fully extended (full-body view). - ChatGPT 4o
ChatGPT 4o 6 / 10 Ranked 7th of 12, this succeeds at depicting a difficult balance but only partially fulfills the extension requirement. Grain and ambiguous overlapping anatomy further separate it from the higher-ranked photographs.
A yoga practitioner in a complex pose, arms and legs fully extended (full-body view). - Imagen 4.0 Ultra
Imagen 4.0 Ultra 8 / 10 This is among the strongest results for exact prompt compliance and readable anatomy. It falls short of the top tier mainly through less refined extremity detail and a more ordinary presentation.
A studio headshot of a young man with one blue eye and one green eye (heterochromia). - ChatGPT 4o
ChatGPT 4o 10 / 10 Ranked 2nd of 13, this is an exceptionally convincing professional headshot with no conspicuous generation artifacts. Its greatest strength is rendering the requested trait clearly while preserving completely believable eye coloration.
A studio headshot of a young man with one blue eye and one green eye (heterochromia). - Imagen 4.0 Ultra
Imagen 4.0 Ultra 6 / 10 Ranked 7th, this is a competent headshot with very clear heterochromia. The polished complexion and exaggerated iris treatment weaken the realism of the specific physical trait being evaluated.
An armchair designed in the shape of an avocado, presented in a contemporary product photography style with professional studio lighting against a minimalist background. - ChatGPT 4o
ChatGPT 4o 8 / 10 Ranks third of 13 for clear prompt fulfillment and a convincing furniture silhouette. Darker rendering and slightly formulaic surface treatment keep it below the strongest examples.
An armchair designed in the shape of an avocado, presented in a contemporary product photography style with professional studio lighting against a minimalist background. - Imagen 4.0 Ultra
Imagen 4.0 Ultra 6 / 10 Rank 7 of 12. The avocado concept is strong, but the missing armchair structure and visibly synthetic surfacing are meaningful shortcomings. It succeeds more as a design visualization than as realistic product photography.
A technology magazine cover featuring the headline "Tech Innovations of 2025," clean modern design style. - ChatGPT 4o
ChatGPT 4o 9 / 10 This is the best typography-first interpretation and belongs among the three strongest results. It sacrifices illustrative originality but delivers outstanding readability, restraint, and magazine credibility.
A technology magazine cover featuring the headline "Tech Innovations of 2025," clean modern design style. - Imagen 4.0 Ultra
Imagen 4.0 Ultra 7 / 10 Rank 7 of 12: accurate typography and a coherent palette make this a solid interpretation. Ambiguous hand construction and a less developed editorial composition hold it back.
Photorealistic close-up portrait of a person clearly performing the American Sign Language gesture for "thank you," hand positioned visibly in front of the chest, clear expression on face, neutral background - ChatGPT 4o
ChatGPT 4o 8 / 10 This is one of the clearest and most anatomically reliable depictions of the requested starting pose. It ranks third because its presentation is less compelling than the warmer leading portraits.
Photorealistic close-up portrait of a person clearly performing the American Sign Language gesture for "thank you," hand positioned visibly in front of the chest, clear expression on face, neutral background - Imagen 4.0 Ultra
Imagen 4.0 Ultra 2 / 10 Rank 12 of 12: this combines the wrong sign with the set's clearest finger-count failure. Its polished facial rendering does not rescue the central task.

More comparisons