overall96
rank#5
credits1
median s47
max res1K
5thOpenAI · text-to-image
1K tiers · 16 reference images
Benchmark v1.0 · 29 October 2026 · 26 models
GPT-Image 2
Latest OpenAI image model at the medium quality tier.
- Best for peopletied
- Best for producttied
- Best for scenetied
- Technical qualitytied
- Aestheticstied
- Text rendering (OCR)tied
0weeks at #1
0weeks in top 3
1week in top 10
#5best rank1 benchmark run · in index since 29 October 2026
Benchmark record 12 prompts · v1.0
| Prompt | Score | Adh | Tech | Aes | Sec | Pos |
|---|---|---|---|---|---|---|
 PeopleStudio portrait of a woman in her forties with silver-streaked hair, soft window light from the left, neutral grey background, 85mm look, natural skin textureExcellent adherence to prompt with high-quality execution and pleasing aesthetics. | 98 | 100 | 95 | 95 | 37 | 1/26 |
 PeopleA barista in a linen apron laughing while pouring latte art, warm cafe interior, shallow depth of field, candid editorial styleExcellent adherence and quality; warm and inviting aesthetic. | 95 | 95 | 95 | 95 | 56.7 | 1/26 |
 ProductA matte black ceramic coffee mug on a walnut table, single soft light from the left, minimal studio backdrop, product photographyExcellent adherence and quality, very professional aesthetic. | 98 | 100 | 95 | 95 | 37.9 | 1/26 |
 ProductA pair of white running shoes floating mid-air over a pastel pink background, even studio lighting, catalogue shotPerfect adherence and high-quality execution. | 98 | 100 | 95 | 95 | 33.7 | 1/26 |
 TextA vintage cinema marquee at dusk with the words OPEN ALL NIGHT in bold red letters, neon glow, wet street reflectionsAccurate depiction with minor extra elements, high quality and aesthetic appeal. | 91 | 88 | 93 | 94 | 49.5 | 19/26 |
 TextA minimalist poster with the single word BALANCE in large serif type centred on cream paper, one small green circle below the wordPerfect adherence with high-quality minimalist design. | 98 | 100 | 95 | 95 | 31.7 | 1/26 |
 IllustrationAnime-style illustration of a young woman with a red umbrella standing on a rainy city street at night, cel shading, vibrant neon reflectionsExcellent adherence to prompt with vibrant neon reflections and cel shading. | 95 | 95 | 95 | 95 | 47 | 8/26 |
 IllustrationFlat vector illustration of a cosy reading nook with a cat asleep on an armchair, limited palette of four colours, clean geometric shapesExcellent adherence to prompt with clean geometric shapes and limited palette. | 94 | 95 | 93 | 94 | 70.9 | 9/26 |
 SceneA modern glass house on a cliff at golden hour, long shadows, calm ocean below, architectural photographyExcellent adherence and quality; stunning architectural photography. | 98 | 100 | 95 | 95 | 46.2 | 1/26 |
 SceneMisty pine forest at dawn with rays of light through the trees and a narrow wooden boardwalk, wide angleBeautifully captures the prompt with excellent light and composition. | 98 | 100 | 95 | 95 | 79.9 | 1/26 |
 CompositionThree red apples and one green pear arranged in a single row on a blue plate, top-down view, white tableclothPerfect adherence and high-quality execution. | 98 | 100 | 95 | 95 | 29.7 | 1/26 |
 CompositionA small orange robot holding a yellow umbrella standing to the left of a tall blue mailbox, overcast day, photorealExcellent adherence and photorealism, slight text issue on mailbox. | 93 | 95 | 90 | 92 | 60 | 16/26 |
Accumulated benchmark v1.0 · price list
12images scored
46.6median seconds
70.9p90 seconds
1credits / image
$0.19≈ usd / image
96.2quality per credit
1Kmax resolution
16reference images
1per request
Five ways of judging rank out of 26
Overall index5th of 26
Prompt adherence6th of 26
Technical quality6th of 26
Aesthetics5th of 26
Text rendering (OCR)1st of 26
Counts & placement1st of 26
Adherence, technical and aesthetics come from a blind vision-model judge; text and counts are objective checks. The overall index blends the first three 50/30/20.
Profile by prompt type axis 90–100
GPT-Image 2Field median (26 models)
Output map every render, unretouched
AI-generated. First result returned, no retries, no editing.
Against the field GPT-Image 2 vs median · axis 90–100
GPT-Image 2Field median
Reading the card
GPT-Image 2 finishes 5th of 26 on the overall index at 96.2.
Its strongest kind of prompt is product (98 against a field median of 96.5); its weakest is illustration (94.5 against 94).
OCR found the exact words in 100% of the text prompts and 100% of the count, colour and placement checks passed.
A render takes a median 46.6 s, 22nd of 26 for speed and slower than the field median of 20.3 s, and costs 1 credit (about $0.19 on the Premium plan), which puts it 2nd of 26 on quality per credit.
Pick it for
- Product98 · 1st of 26 · clean studio lighting, material fidelity, catalogue framing
- Scene98 · 1st of 26 · landscape and architecture, depth, light direction
- People96.5 · 1st of 26 · skin, hands, eyes and natural light on a human subject
- Value96.2 quality points per credit, 2nd of 26
Weigh up
- Illustration94.5 · 8th of 26 · holding a named art style with clean shapes and palette
- Speed46.6 s median is slower than the field’s 20.3 s
Nearest rivals closest overall scores
In the studio what you can set
- Vendor
- OpenAI
- Resolution tiers
- 1K
- Highest output
- 1K
- Reference images
- Up to 16
- Images per request
- 1
- Seed
- No
- Negative prompt
- No
- Base price
- 1 credit · ≈ $0.19
Questions about GPT-Image 2
- Is GPT-Image 2 good at rendering text in images?
- In benchmark v1.0 the OCR check found the exact requested phrase in 100% of GPT-Image 2's text-prompt renders, and the blind judge gave its text category 94.5 out of 100 (field median 95.5). Every word came out exactly as written.
- How fast is GPT-Image 2?
- Median 46.6 seconds from submit to finished image across the 12 benchmark prompts, with the slowest tenth at 70.9 seconds. That is 22nd of 26 models; the field median is 20.3 seconds.
- How much does GPT-Image 2 cost per image?
- 1 credit at the base 1K tier, about $0.19 at the Premium plan rate of $0.19 per credit. Higher resolution tiers can cost more; see pricing.
- GPT-Image 2 or Nano Banana 2.1?
- Nano Banana 2.1 scores 96.2 overall against 96.2 for GPT-Image 2, inside the judge’s noise, so treat them as equal on quality and decide on speed and price. GPT-Image 2 renders in 46.6 s for 1 credit; Nano Banana 2.1 in 35 s for 1 credit.
- What is GPT-Image 2 best used for?
- It ranks in the top three for people, product, scene prompts in this run. Latest OpenAI image model at the medium quality tier.
- How long has GPT-Image 2 been at the top?
- GPT-Image 2 is currently 5th; its best rank across 1 benchmark run is #5, and it has been in the top 10 for 1 week.
- How were these numbers measured?
- Every model rendered the same 12 prompts at matched settings. A vision-model judge scored each image blind on adherence, technical quality and aesthetics (blended 50/30/20 into the overall index); OCR and fixed yes/no checks measured text accuracy and composition without a judge. Seconds were timed through the production API. Benchmark v1.0, 29 October 2026.