whatmodel.sh / image / GPT Image 1
GPT Image 1
↗ share
+ contribute
GPT Image 1 is a closed-weights image model from OpenAI, released April 2025. Licence: proprietary.
Score within the LMArena text-to-image comparison set only.
$ how it compares
The image models ranked either side of GPT Image 1, out of 197 with a score, then more from OpenAI.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 40 |  Wan 2.7 Image | 62±2 | api $0.030/image | ||
| 41 |  Imagen 4 Ultra | 61±2 | api $0.060/image | ||
| 42 | GPT Image 1 | 61±1 | api $0.042/image | ||
| 43 | Wan 2.6 Text-to-Image | 61±3 | api $0.030/image | ||
| 44 |  Reve V1.1 | 61±4 | comparison only | ||
| more from OpenAI | |||||
| 1 | GPT Image 2.5 Sunburst | 92±2 | api $0.013/image | ||
| 2 | GPT Image 2.5 Flare | 88±2 | api $0.013/image | ||
| 3 | GPT Image 2 | 85±1 | api $0.053/image | ||
$ what it's good at
text to image▂▅▅▃▄▄█▃▁▂▁▁ #33 of 126 · 75% of its score · 11 tests
Beats Wan 2.7 Image Pro, Wan 2.7 Image, Imagen 4 Ultra here, though they rank higher overall. Trails ERNIE-Image, ERNIE-Image-Turbo here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GenExam (relaxed overall, GPT-5 judge) | 62.2% ✎ | ▄█▂▂▂▂▆▄·▂·▂ beats 14/16 | 26% |
| LMArena text-to-image | 1115.6 ✎ | ▂▂▃▄▇█▄▂▄▂·▂ beats 32/75 | 16% |
| ServImage task acceptance (human payment decisions, commercial design tasks) | 48.2% ✎ | ▂▃··▂··▃█··▃ beats 11/11 | 7.6% |
| T2I-ReasonBench (overall) | 86.3% ✎ | ▅▂█▂▂▂····▃▂ beats 12/13 | 5.4% |
| UltraText Bench (Composite, bilingual dense text rendering) | 37.9% ✎ | ▁▂▂█▃▁▂▃·▁·▁ beats 5/22 | 4.4% |
| TIIF-Bench v3 (short prompts, TIIF-Evaluator) | 86.1% ✎ | ▂█····▄▄▄··▂ beats 6/11 | 4.2% |
| GenEval | 80.8% ✎ | ██·▄·▄·█···▄ beats 8/8 | 3.6% |
| ImagenWorld text-guided generation (human overall) | 91.0% ✎ | ▄··█▄·▄··▄·▄ beats 6/6 | 2.6% |
3 more tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| MMMG image generation (average over 15 tasks) | 70.7% ✎ | █··█▄▄·▄▄·▄▄ beats 9/9 | 2.6% |
| Qwen-Image-Bench (Q-Judger, v1) | 54.1% ✎ | █▅▅·▅▅█·█··▃ beats 7/17 | 1.8% |
| GenAI-Bench (VQAScore overall) | 85.0% ✎ | █···█·█····█ beats 3/3 | 0.5% |
image to image / editing▁▁·▃▂▃█▃▃▄▁▁ #32 of 62 · 29% of its score · 8 tests
Beats FLUX.2 [flex] here, though they rank higher overall. Trails Qwen-Image-Edit-2509, Kling Image O1, Seedream 4.0 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| ServImage task acceptance (human payment decisions, commercial design tasks) | 48.2% ✎ | ▂▃··▂··▃█··▃ beats 11/11 | 7.6% |
| Arena image edit | 1138.7 ✎ | ▂▁▁▃▆▇█▄▆▅▁▂ beats 4/53 | 6.3% |
| CEdit-Bench-CN overall (G_O, GPT-4o judge) | 82.1% ✎ | ▃·········█▃ beats 7/7 | 4.1% |
| GEBench GUI state generation (English GE-Score) | 55.6% ✎ | ▄▄····▄▄█▄·█ beats 6/8 | 3.2% |
| CEdit-Bench-EN overall (G_O, GPT-4o judge) | 81.7% ✎ | ▄·▄···▄██··▄ beats 7/7 | 2.8% |
| UnicBench (EN, overall) | 83.5% ✎ | ▂····▂▄▂▄·█▃ beats 11/11 | 2.3% |
| CREval-Bench creative image editing (Overall, GPT-4o QA evaluator) | 79.0% ✎ | ▄█▄··█··▄·▄▄ beats 6/8 | 1.5% |
| ImagenWorld text-guided editing (human overall) | 82.0% ✎ | █··█··██···█ beats 4/4 | 1.1% |
other tests2 not in a task · 3.3% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| EdiVal multi-turn editing overall (EdiVal-O after turn 3) | 53.8% ✎ | ▂·▂▃·▂·▃▂·▃█ beats 9/14 | 2.3% |
| ViGoR-Bench zero-shot visual reasoning, image models (Result Avg) | 46.9% ✎ | ▂···▂·▂█▆▅·▂ beats 13/15 | 1.0% |
$ where to run it
Through an API
api · $0.042/image · developers.openai.com ↗ · price sourceUSD for one generated 1024×1024 medium-quality image. Output-only published price; input text/image token charges are additional. Scheduled deprecation 2026-10-23. Product: Image generation API. API model: gpt-image-1. Settings: quality=medium, width=1024, height=1024, n=1. Price basis: generated image output component. Additional/excluded charges: input text/image tokens. Currency: USD.
$ how the score was worked out · GenExam: A Multidisciplinary… 26%, LMArena leaderboard dataset 23%, ServImage: An Image Generatio… 8% and 12 more
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json