overall97
rank#2
credits3
median s43
max res1.5K
2ndOpenAI · text-to-image
1.5K tiers · 10 reference images
Benchmark v1.0 · 29 October 2026 · 26 models
GPT-Image 1.5
OpenAI model at high quality. Strong instruction following.
- Best for illustration
- Best overalltied
- Best for scenetied
- Technical qualitytied
- Aestheticstied
- Text rendering (OCR)tied
0weeks at #1
1week in top 3since 29 October 2026
1week in top 10
#2best 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 and quality with pleasing aesthetics. | 95 | 95 | 95 | 95 | 42.6 | 6/26 |
 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 candid style captured well. | 95 | 95 | 95 | 95 | 59.9 | 1/26 |
 ProductA matte black ceramic coffee mug on a walnut table, single soft light from the left, minimal studio backdrop, product photographyPerfect execution of prompt with excellent lighting and composition. | 98 | 100 | 95 | 95 | 29.4 | 1/26 |
 ProductA pair of white running shoes floating mid-air over a pastel pink background, even studio lighting, catalogue shotExcellent adherence and quality, visually appealing catalogue shot. | 95 | 95 | 95 | 95 | 73.7 | 12/26 |
 TextA vintage cinema marquee at dusk with the words OPEN ALL NIGHT in bold red letters, neon glow, wet street reflectionsExcellent depiction with minor text discrepancies. | 94 | 93 | 95 | 95 | 170.3 | 13/26 |
 TextA minimalist poster with the single word BALANCE in large serif type centred on cream paper, one small green circle below the wordPerfect adherence to prompt with high-quality execution. | 98 | 100 | 95 | 95 | 85.6 | 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 with vibrant neon reflections and cel shading. | 98 | 100 | 95 | 95 | 40.1 | 1/26 |
 IllustrationFlat vector illustration of a cosy reading nook with a cat asleep on an armchair, limited palette of four colours, clean geometric shapesPerfect adherence with clean execution and pleasing aesthetics. | 98 | 100 | 95 | 95 | 28.9 | 1/26 |
 SceneA modern glass house on a cliff at golden hour, long shadows, calm ocean below, architectural photographyExcellent adherence and quality; stunning architectural shot. | 98 | 100 | 95 | 95 | 42.5 | 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 | 36.5 | 1/26 |
 CompositionThree red apples and one green pear arranged in a single row on a blue plate, top-down view, white tableclothPerfect adherence with high quality and aesthetics. | 98 | 100 | 95 | 95 | 31.5 | 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, minor quality issues. | 93 | 95 | 90 | 92 | 59.9 | 16/26 |
Accumulated benchmark v1.0 · price list
12images scored
42.6median seconds
85.6p90 seconds
3credits / image
$0.57≈ usd / image
32.2quality per credit
1.5Kmax resolution
10reference images
1per request
Five ways of judging rank out of 26
Overall index2nd of 26
Prompt adherence2nd of 26
Technical quality2nd of 26
Aesthetics2nd 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 1.5Field median (26 models)
Output map every render, unretouched
AI-generated. First result returned, no retries, no editing.
Against the field GPT-Image 1.5 vs median · axis 90–100
GPT-Image 1.5Field median
Reading the card
GPT-Image 1.5 finishes 2nd of 26 on the overall index at 96.5, within 1 point of Seedream 5.0 Pro and therefore a statistical tie for first.
Its strongest kind of prompt is scene (98 against a field median of 96.5); its weakest is people (95 against 95).
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 42.6 s, 21st of 26 for speed and slower than the field median of 20.3 s, and costs 3 credits (about $0.57 on the Premium plan), which puts it 23rd of 26 on quality per credit.
Pick it for
- Illustration98 · 1st of 26 · holding a named art style with clean shapes and palette
- Scene98 · 1st of 26 · landscape and architecture, depth, light direction
- Composition95.5 · 10th of 26 · object counts, colours and left/right placement
Weigh up
- People95 · 6th of 26 · skin, hands, eyes and natural light on a human subject
- Speed42.6 s median is slower than the field’s 20.3 s
- Price3 credits per image, 23rd of 26 on price; 16 models in the index cost one
Nearest rivals closest overall scores
In the studio what you can set
- Vendor
- OpenAI
- Resolution tiers
- 1.5K
- Highest output
- 1.5K
- Reference images
- Up to 10
- Images per request
- 1
- Seed
- No
- Negative prompt
- No
- Base price
- 3 credits · ≈ $0.57
Questions about GPT-Image 1.5
- Is GPT-Image 1.5 good at rendering text in images?
- In benchmark v1.0 the OCR check found the exact requested phrase in 100% of GPT-Image 1.5's text-prompt renders, and the blind judge gave its text category 96 out of 100 (field median 95.5). Every word came out exactly as written.
- How fast is GPT-Image 1.5?
- Median 42.6 seconds from submit to finished image across the 12 benchmark prompts, with the slowest tenth at 85.6 seconds. That is 21st of 26 models; the field median is 20.3 seconds.
- How much does GPT-Image 1.5 cost per image?
- 3 credits at the base 1.5K tier, about $0.57 at the Premium plan rate of $0.19 per credit. Higher resolution tiers can cost more; see pricing.
- GPT-Image 1.5 or Nano Banana Pro?
- Nano Banana Pro scores 96.5 overall against 96.5 for GPT-Image 1.5, inside the judge’s noise, so treat them as equal on quality and decide on speed and price. GPT-Image 1.5 renders in 42.6 s for 3 credits; Nano Banana Pro in 47.7 s for 3 credits.
- What is GPT-Image 1.5 best used for?
- It ranks in the top three for illustration, scene prompts in this run. OpenAI model at high quality. Strong instruction following.
- How long has GPT-Image 1.5 been at the top?
- GPT-Image 1.5 has been in the top 3 for 1 week, since the 29 October 2026 run; its best rank so far is #2. Weeks are counted from the first benchmark run that placed it there.
- 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.