PRISM-Bench (image generation)
Models generate images from English or Chinese prompts. Vision-language judges score visual appeal and prompt alignment across seven tracks, including composition, lettering, and style.
The authors also released FLUX-Reason-6M for training. The evaluation prompts and code are in the benchmark repository.
- Released
- September 2025
- Kind
- Image generation and editing
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Image, video and 3D generation
- Last model added
- September 2025
- Scores refreshed
- November 20th, 2026
GPT-Image-1 leads with 86.3 among models released by September 2025 GPT-Image-1 leads with 86.3 among models released by September 2025
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in September 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 86.3 GPT-Image-1
- 85.3
Gemini 2.5 Flash Image Preview (nano-banana)
- 79.9
Qwen-Image
- 79.6
Seedream 3.0
- 75.9 HiDream-I1-Full
- 74.7 FLUX.1-Krea-dev
- 73.7 FLUX.1 [dev]
- 73.7
Stable Diffusion 3.5 Large
- 70.2 HiDream-I1-Dev
- 69.5 SD3.5-Medium
- 66.1 SD3-Medium
- 65.2
BAGEL
- 64.7 FLUX.1-schnell
- 64.1 Playground
- 62.1
Janus-Pro-7B
- 60.4 SDXL
- 50.8 SD2.1
- 46.0 SD1.5
- 1 GPT-Image-1 GPT4.1, High 86.3
- 2
Gemini 2.5 Flash Image Preview (nano-banana) GPT4.1 85.3
- 3
Qwen-Image GPT4.1 79.9
- 4
Seedream 3.0 GPT4.1 79.6
- 5 HiDream-I1-Full GPT4.1 75.9
- 6 FLUX.1-Krea-dev GPT4.1 74.7
- 7 FLUX.1 [dev] GPT4.1 73.7
- 8
Stable Diffusion 3.5 Large GPT4.1 73.7
- 9 HiDream-I1-Dev GPT4.1 70.2
- 10 SD3.5-Medium GPT4.1 69.5
- 11 SD3-Medium Qwen2.5-VL 66.1
- 12
BAGEL GPT4.1, CoT 65.2
- 13 FLUX.1-schnell Qwen2.5-VL 64.7
- 14 Playground GPT4.1 64.1
- 15
Janus-Pro-7B Qwen2.5-VL 62.1
- 16 SDXL GPT4.1 60.4
- 17 SD2.1 Qwen2.5-VL 50.8
- 18 SD1.5 Qwen2.5-VL 46.0
Source: Official PRISM-Bench (image generation) leaderboard
- 1 GPT-Image-1 GPT4.1, High 86.3
- 2
Gemini 2.5 Flash Image Preview (nano-banana) GPT4.1 85.3
- 3
Qwen-Image GPT4.1 79.9
- 4
Seedream 3.0 GPT4.1 79.6
- 5 FLUX.1 [dev] GPT4.1 73.7
- 6
Stable Diffusion 3.5 Large GPT4.1 73.7
- 7
BAGEL GPT4.1, CoT 65.2
- 8
Janus-Pro-7B Qwen2.5-VL 62.1
Source: Official PRISM-Bench (image generation) leaderboard
| # | Entry | Score | Setting | Released | Run on | Link | Source |
|---|---|---|---|---|---|---|---|
| 1 | GPT-Image-1 | 86.3 | GPT4.1, High | Sep 2025 | Sep 2025 | platform.openai.com | Official leaderboard |
| 2 | | 85.3 | GPT4.1 | Sep 2025 | Sep 2025 | deepmind.google | Official leaderboard |
| 3 | | 79.9 | GPT4.1 | Sep 2025 | Sep 2025 | huggingface.co | Official leaderboard |
| 4 | | 79.6 | GPT4.1 | Sep 2025 | Sep 2025 | seed.bytedance.com | Official leaderboard |
| 5 | HiDream-I1-Full | 75.9 | GPT4.1 | Sep 2025 | huggingface.co | Official leaderboard | |
| 6 | FLUX.1-Krea-dev | 74.7 | GPT4.1 | Sep 2025 | huggingface.co | Official leaderboard | |
| 7 | FLUX.1 [dev] | 73.7 | GPT4.1 | Sep 2025 | Sep 2025 | huggingface.co | Official leaderboard |
| 8 | | 73.7 | GPT4.1 | Sep 2025 | Sep 2025 | huggingface.co | Official leaderboard |
| 9 | HiDream-I1-Dev | 70.2 | GPT4.1 | Sep 2025 | huggingface.co | Official leaderboard | |
| 10 | SD3.5-Medium | 69.5 | GPT4.1 | Sep 2025 | huggingface.co | Official leaderboard | |
| 11 | SD3-Medium | 66.1 | Qwen2.5-VL | Sep 2025 | huggingface.co | Official leaderboard | |
| 12 | | 65.2 | GPT4.1, CoT | Sep 2025 | Sep 2025 | github.com | Official leaderboard |
| 13 | FLUX.1-schnell | 64.7 | Qwen2.5-VL | Sep 2025 | huggingface.co | Official leaderboard | |
| 14 | Playground | 64.1 | GPT4.1 | Sep 2025 | huggingface.co | Official leaderboard | |
| 15 | | 62.1 | Qwen2.5-VL | Jan 2025 | Sep 2025 | huggingface.co | Official leaderboard |
| 16 | SDXL | 60.4 | GPT4.1 | Sep 2025 | huggingface.co | Official leaderboard | |
| 17 | SD2.1 | 50.8 | Qwen2.5-VL | Sep 2025 | huggingface.co | Official leaderboard | |
| 18 | SD1.5 | 46.0 | Qwen2.5-VL | Sep 2025 | huggingface.co | Official leaderboard |
| # | Model | Score | Setting | Released | Run on | Link | Source |
|---|---|---|---|---|---|---|---|
| 1 | GPT-Image-1 | 86.3 | GPT4.1, High | Sep 2025 | Sep 2025 | platform.openai.com | Official leaderboard |
| 2 | | 85.3 | GPT4.1 | Sep 2025 | Sep 2025 | deepmind.google | Official leaderboard |
| 3 | | 79.9 | GPT4.1 | Sep 2025 | Sep 2025 | huggingface.co | Official leaderboard |
| 4 | | 79.6 | GPT4.1 | Sep 2025 | Sep 2025 | seed.bytedance.com | Official leaderboard |
| 5 | FLUX.1 [dev] | 73.7 | GPT4.1 | Sep 2025 | Sep 2025 | huggingface.co | Official leaderboard |
| 6 | | 73.7 | GPT4.1 | Sep 2025 | Sep 2025 | huggingface.co | Official leaderboard |
| 7 | | 65.2 | GPT4.1, CoT | Sep 2025 | Sep 2025 | github.com | Official leaderboard |
| 8 | | 62.1 | Qwen2.5-VL | Jan 2025 | Sep 2025 | huggingface.co | Official leaderboard |
Rankings by part
We read 47 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Overall · Align, Overall · Aes and Imagination · Align. GPT-Image-1 leads 36 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Overall · Align
8 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑1 87.1
- 2 GPT-Image-1 ↓1 86.9
- 3
Qwen-Image 81.1
- 4
Seedream 3.0 80.5
- 5
Stable Diffusion 3.5 Large ↑1 73.9
- 6 FLUX.1 [dev] ↓1 72.4
- 7
BAGEL 67.5
- 8
Janus-Pro-7B 64.9
Overall · Aes
8 models we track have a score here.
- 1 GPT-Image-1 85.6
- 2
Gemini 2.5 Flash Image Preview (nano-banana) 83.4
- 3
Seedream 3.0 ↑1 78.7
- 4
Qwen-Image ↓1 78.6
- 5 FLUX.1 [dev] 74.9
- 6
Stable Diffusion 3.5 Large 73.5
- 7
BAGEL 65.0
- 8
Janus-Pro-7B 59.4
Imagination · Align
8 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑1 92.4
- 2 GPT-Image-1 ↓1 86.2
- 3
Qwen-Image 80.5
- 4
Seedream 3.0 77.3
- 5
Stable Diffusion 3.5 Large ↑1 73.3
- 6
Janus-Pro-7B ↑2 70.4
- 7
BAGEL 69.4
- 8 FLUX.1 [dev] ↓3 68.1
Imagination · Aes
8 models we track have a score here.
- 1 GPT-Image-1 86.6
- 2
Gemini 2.5 Flash Image Preview (nano-banana) 84.8
- 3
Qwen-Image 78.6
- 4
Seedream 3.0 76.4
- 5
BAGEL ↑2 74.2
- 6 FLUX.1 [dev] ↓1 74.0
- 7
Stable Diffusion 3.5 Large ↓1 71.2
- 8
Janus-Pro-7B 65.8
Imagination · Avg
8 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑1 88.6
- 2 GPT-Image-1 ↓1 86.4
- 3
Qwen-Image 79.6
- 4
Seedream 3.0 76.9
- 5
Stable Diffusion 3.5 Large ↑1 72.3
- 6
BAGEL ↑1 71.3
- 7 FLUX.1 [dev] ↓2 71.1
- 8
Janus-Pro-7B 68.1
Entity · Align
8 models we track have a score here.
- 1 GPT-Image-1 90.0
- 2
Gemini 2.5 Flash Image Preview (nano-banana) 87.0
- 3
Seedream 3.0 ↑1 81.3
- 4
Qwen-Image ↓1 79.5
- 5
Stable Diffusion 3.5 Large ↑1 76.8
- 6 FLUX.1 [dev] ↓1 70.7
- 7
Janus-Pro-7B ↑1 68.6
- 8
BAGEL ↓1 67.6
Entity · Aes
8 models we track have a score here.
- 1 GPT-Image-1 86.3
- 2
Gemini 2.5 Flash Image Preview (nano-banana) 81.3
- 3
Seedream 3.0 ↑1 74.2
- 4
Qwen-Image ↓1 73.2
- 5
Stable Diffusion 3.5 Large ↑1 72.7
- 6 FLUX.1 [dev] ↓1 71.2
- 7
Janus-Pro-7B ↑1 63.5
- 8
BAGEL ↓1 60.0
Entity · Avg
8 models we track have a score here.
- 1 GPT-Image-1 88.2
- 2
Gemini 2.5 Flash Image Preview (nano-banana) 84.2
- 3
Seedream 3.0 ↑1 77.7
- 4
Qwen-Image ↓1 76.3
- 5
Stable Diffusion 3.5 Large ↑1 74.8
- 6 FLUX.1 [dev] ↓1 71.0
- 7
Janus-Pro-7B ↑1 66.0
- 8
BAGEL ↓1 61.2
Text rendering · Align
8 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑1 72.8
- 2 GPT-Image-1 ↓1 68.8
- 3
Seedream 3.0 ↑1 58.8
- 4
Qwen-Image ↓1 57.9
- 5
Stable Diffusion 3.5 Large ↑1 53.6
- 6 FLUX.1 [dev] ↓1 52.3
- 7
BAGEL 30.2
- 8
Janus-Pro-7B 23.1
Text rendering · Aes
8 models we track have a score here.
- 1 GPT-Image-1 86.8
- 2
Gemini 2.5 Flash Image Preview (nano-banana) 84.3
- 3
Seedream 3.0 ↑1 74.0
- 4
Stable Diffusion 3.5 Large ↑2 73.1
- 5 FLUX.1 [dev] 73.0
- 6
Qwen-Image ↓3 71.2
- 7
Janus-Pro-7B ↑1 50.3
- 8
BAGEL ↓1 44.5
Text rendering · Avg
8 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑1 78.5
- 2 GPT-Image-1 ↓1 76.8
- 3
Seedream 3.0 ↑1 66.4
- 4
Qwen-Image ↓1 64.5
- 5
Stable Diffusion 3.5 Large ↑1 63.3
- 6 FLUX.1 [dev] ↓1 62.6
- 7
BAGEL 37.4
- 8
Janus-Pro-7B 36.7
Style · Align
8 models we track have a score here.
- 1 GPT-Image-1 92.8
- 2
Gemini 2.5 Flash Image Preview (nano-banana) 90.5
- 3
Qwen-Image 86.6
- 4
Seedream 3.0 84.4
- 5
Stable Diffusion 3.5 Large ↑1 77.3
- 6 FLUX.1 [dev] ↓1 72.6
- 7
Janus-Pro-7B ↑1 71.4
- 8
BAGEL ↓1 69.0
Style · Aes
8 models we track have a score here.
- 1 GPT-Image-1 93.3
- 2
Gemini 2.5 Flash Image Preview (nano-banana) 90.8
- 3
Qwen-Image 88.7
- 4
Seedream 3.0 87.4
- 5
Stable Diffusion 3.5 Large ↑1 84.2
- 6 FLUX.1 [dev] ↓1 80.5
- 7
Janus-Pro-7B ↑1 75.2
- 8
BAGEL ↓1 71.3
Style · Avg
8 models we track have a score here.
- 1 GPT-Image-1 93.1
- 2
Gemini 2.5 Flash Image Preview (nano-banana) 90.7
- 3
Qwen-Image 86.6
- 4
Seedream 3.0 85.7
- 5
Stable Diffusion 3.5 Large ↑1 80.7
- 6 FLUX.1 [dev] ↓1 76.4
- 7
Janus-Pro-7B ↑1 72.9
- 8
BAGEL ↓1 69.6
Affection · Align
8 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑1 96.0
- 2
Qwen-Image ↑1 91.6
- 3 GPT-Image-1 ↓2 90.7
- 4
Seedream 3.0 90.5
- 5 FLUX.1 [dev] 88.3
- 6
BAGEL ↑1 87.1
- 7
Stable Diffusion 3.5 Large ↓1 87.1
- 8
Janus-Pro-7B 80.7
Affection · Aes
8 models we track have a score here.
- 1 FLUX.1 [dev] ↑4 91.1
- 2 GPT-Image-1 ↓1 90.9
- 3
Seedream 3.0 ↑1 90.3
- 4
Qwen-Image ↓1 89.1
- 5
Gemini 2.5 Flash Image Preview (nano-banana) ↓3 88.2
- 6
Stable Diffusion 3.5 Large 85.2
- 7
BAGEL 81.4
- 8
Janus-Pro-7B 71.5
Affection · Avg
8 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑1 92.1
- 2 GPT-Image-1 ↓1 90.8
- 3
Qwen-Image 90.4
- 4
Seedream 3.0 89.8
- 5 FLUX.1 [dev] 89.7
- 6
Stable Diffusion 3.5 Large 86.2
- 7
BAGEL 83.8
- 8
Janus-Pro-7B 75.4
Composition · Align
8 models we track have a score here.
- 1 GPT-Image-1 96.2
- 2
Qwen-Image ↑1 93.9
- 3
Seedream 3.0 ↑1 93.6
- 4
Gemini 2.5 Flash Image Preview (nano-banana) ↓2 92.5
- 5
BAGEL ↑2 90.5
- 6 FLUX.1 [dev] ↓1 89.0
- 7
Stable Diffusion 3.5 Large ↓1 87.8
- 8
Janus-Pro-7B 83.7
Composition · Aes
8 models we track have a score here.
- 1 GPT-Image-1 89.4
- 2
Gemini 2.5 Flash Image Preview (nano-banana) 88.5
- 3
Qwen-Image 86.9
- 4
Seedream 3.0 86.3
- 5
Stable Diffusion 3.5 Large ↑1 84.7
- 6 FLUX.1 [dev] ↓1 84.6
- 7
BAGEL 77.9
- 8
Janus-Pro-7B 71.1
Composition · Avg
8 models we track have a score here.
- 1 GPT-Image-1 92.8
- 2
Gemini 2.5 Flash Image Preview (nano-banana) 90.5
- 3
Qwen-Image 90.3
- 4
Seedream 3.0 89.8
- 5 FLUX.1 [dev] 86.8
- 6
Stable Diffusion 3.5 Large 85.9
- 7
BAGEL 83.2
- 8
Janus-Pro-7B 76.7
Long text · Align
8 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑1 85.9
- 2 GPT-Image-1 ↓1 83.8
- 3
Qwen-Image 83.8
- 4
Seedream 3.0 83.2
- 5 FLUX.1 [dev] 70.6
- 6
BAGEL ↑1 68.1
- 7
Stable Diffusion 3.5 Large ↓1 65.8
- 8
Janus-Pro-7B 63.9
Long text · Aes
8 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑1 80.6
- 2 GPT-Image-1 ↓1 77.5
- 3
Seedream 3.0 ↑1 76.4
- 4
Qwen-Image ↓1 70.9
- 5 FLUX.1 [dev] 58.5
- 6
BAGEL ↑1 55.3
- 7
Stable Diffusion 3.5 Large ↓1 52.2
- 8
Janus-Pro-7B 49.0
Long text · Avg
8 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑1 81.1
- 2 GPT-Image-1 ↓1 78.3
- 3
Seedream 3.0 ↑1 76.3
- 4
Qwen-Image ↓1 74.5
- 5 FLUX.1 [dev] 64.6
- 6
BAGEL ↑1 61.7
- 7
Stable Diffusion 3.5 Large ↓1 59.0
- 8
Janus-Pro-7B 56.4
PRISM-Bench-ZH · Overall · Align
4 models we track have a score here.
- 1 GPT-Image-1 87.7
- 2
Seedream 3.0 ↑2 81.9
- 3
Qwen-Image 80.8
- 4
BAGEL ↑3 65.5
PRISM-Bench-ZH · Overall · Aes
4 models we track have a score here.
- 1 GPT-Image-1 87.2
- 2
Seedream 3.0 ↑2 82.0
- 3
Qwen-Image 81.3
- 4
BAGEL ↑3 65.2
PRISM-Bench-ZH · Overall · Avg
4 models we track have a score here.
- 1 GPT-Image-1 87.5
- 2
Seedream 3.0 ↑2 82.0
- 3
Qwen-Image 81.1
- 4
BAGEL ↑3 65.4
PRISM-Bench-ZH · Imagination · Align
4 models we track have a score here.
- 1 GPT-Image-1 88.8
- 2
Qwen-Image ↑1 80.1
- 3
Seedream 3.0 ↑1 77.2
- 4
BAGEL ↑3 75.1
PRISM-Bench-ZH · Imagination · Aes
4 models we track have a score here.
- 1 GPT-Image-1 90.4
- 2
Qwen-Image ↑1 79.6
- 3
Seedream 3.0 ↑1 77.8
- 4
BAGEL ↑3 69.3
PRISM-Bench-ZH · Imagination · Avg
4 models we track have a score here.
- 1 GPT-Image-1 89.6
- 2
Qwen-Image ↑1 79.9
- 3
Seedream 3.0 ↑1 77.5
- 4
BAGEL ↑3 72.2
PRISM-Bench-ZH · Entity · Align
4 models we track have a score here.
- 1 GPT-Image-1 85.9
- 2
Seedream 3.0 ↑2 77.6
- 3
Qwen-Image 75.6
- 4
BAGEL ↑3 62.7
PRISM-Bench-ZH · Entity · Aes
4 models we track have a score here.
- 1 GPT-Image-1 92.4
- 2
Qwen-Image ↑1 79.7
- 3
Seedream 3.0 ↑1 78.6
- 4
BAGEL ↑3 62.2
PRISM-Bench-ZH · Entity · Avg
4 models we track have a score here.
- 1 GPT-Image-1 89.2
- 2
Seedream 3.0 ↑2 78.1
- 3
Qwen-Image 77.7
- 4
BAGEL ↑3 59.1
PRISM-Bench-ZH · Text rendering · Align
4 models we track have a score here.
- 1 GPT-Image-1 83.9
- 2
Seedream 3.0 ↑2 79.7
- 3
Qwen-Image 76.9
- 4
BAGEL ↑3 49.2
PRISM-Bench-ZH · Text rendering · Aes
4 models we track have a score here.
- 1 GPT-Image-1 89.9
- 2
Seedream 3.0 ↑2 88.0
- 3
Qwen-Image 73.1
- 4
BAGEL ↑3 51.9
PRISM-Bench-ZH · Text rendering · Avg
4 models we track have a score here.
- 1 GPT-Image-1 81.5
- 2
Seedream 3.0 ↑2 79.4
- 3
Qwen-Image 69.9
- 4
BAGEL ↑3 39.1
PRISM-Bench-ZH · Style · Align
4 models we track have a score here.
- 1 GPT-Image-1 93.9
- 2
Qwen-Image ↑1 90.2
- 3
Seedream 3.0 ↑1 87.8
- 4
BAGEL ↑3 73.9
PRISM-Bench-ZH · Style · Aes
4 models we track have a score here.
- 1 GPT-Image-1 92.4
- 2
Seedream 3.0 ↑2 88.0
- 3
Qwen-Image 84.3
- 4
BAGEL ↑3 76.7
PRISM-Bench-ZH · Style · Avg
4 models we track have a score here.
- 1 GPT-Image-1 92.8
- 2
Qwen-Image ↑1 87.3
- 3
Seedream 3.0 ↑1 85.5
- 4
BAGEL ↑3 71.3
PRISM-Bench-ZH · Affection · Align
4 models we track have a score here.
- 1 GPT-Image-1 91.5
- 2
Seedream 3.0 ↑2 88.7
- 3
Qwen-Image 87.4
- 4
BAGEL ↑3 81.4
PRISM-Bench-ZH · Affection · Aes
4 models we track have a score here.
- 1 GPT-Image-1 86.5
- 2
Seedream 3.0 ↑2 85.1
- 3
Qwen-Image 84.9
- 4
BAGEL ↑3 78.0
PRISM-Bench-ZH · Affection · Avg
4 models we track have a score here.
- 1 GPT-Image-1 89.0
- 2
Seedream 3.0 ↑2 86.9
- 3
Qwen-Image 86.2
- 4
BAGEL ↑3 79.6
PRISM-Bench-ZH · Composition · Align
4 models we track have a score here.
- 1 GPT-Image-1 92.4
- 2
Seedream 3.0 ↑2 90.3
- 3
Qwen-Image 89.8
- 4
BAGEL ↑3 81.3
PRISM-Bench-ZH · Composition · Aes
4 models we track have a score here.
- 1 GPT-Image-1 97.3
- 2
Seedream 3.0 ↑2 94.4
- 3
Qwen-Image 93.4
- 4
BAGEL ↑3 89.8
PRISM-Bench-ZH · Composition · Avg
4 models we track have a score here.
- 1 GPT-Image-1 94.9
- 2
Seedream 3.0 ↑2 91.1
- 3
Qwen-Image 90.0
- 4
BAGEL ↑3 79.4
PRISM-Bench-ZH · Long text · Align
4 models we track have a score here.
- 1 GPT-Image-1 77.2
- 2
Seedream 3.0 ↑2 74.3
- 3
Qwen-Image 72.6
- 4
BAGEL ↑3 62.4
PRISM-Bench-ZH · Long text · Aes
4 models we track have a score here.
- 1 GPT-Image-1 84.3
- 2
Qwen-Image ↑1 84.2
- 3
Seedream 3.0 ↑1 82.7
- 4
BAGEL ↑3 68.7
PRISM-Bench-ZH · Long text · Avg
4 models we track have a score here.
- 1 GPT-Image-1 80.8
- 2
Seedream 3.0 ↑2 78.5
- 3
Qwen-Image 76.6
- 4
BAGEL ↑3 63.4
More effort adds a median 6 score
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official PRISM-Bench (image generation) leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| GPT-Image-1 | GPT4.1, High | 86.3 | |||
| GPT4.1 | 85.3 | ||||
| GPT-Image-1 | Qwen2.5-VL, High | 80.7 | |||
| Qwen2.5-VL | 80.4 | ||||
| GPT4.1 | 79.9 | ||||
| GPT4.1 | 79.6 | ||||
| Qwen2.5-VL | 76.2 | ||||
| Qwen2.5-VL | 74.1 | ||||
| FLUX.1 [dev] | GPT4.1 | 73.7 | |||
| GPT4.1 | 73.7 | ||||
| Qwen2.5-VL | 70.6 | ||||
| FLUX.1 [dev] | Qwen2.5-VL | 68.5 | |||
| GPT4.1, CoT | 65.2 | ||||
| GPT4.1 | 65.1 | ||||
| Qwen2.5-VL | 62.1 | ||||
| Qwen2.5-VL, CoT | 62.0 | ||||
| Qwen2.5-VL | 62.0 | ||||
| GPT4.1 | 60.7 |
More effort adds a median 6 score
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official PRISM-Bench (image generation) leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| GPT-Image-1 | GPT4.1, High | 86.3 | |||
| GPT4.1 | 85.3 | ||||
| GPT-Image-1 | Qwen2.5-VL, High | 80.7 | |||
| Qwen2.5-VL | 80.4 | ||||
| GPT4.1 | 79.9 | ||||
| GPT4.1 | 79.6 | ||||
| Qwen2.5-VL | 76.2 | ||||
| Qwen2.5-VL | 74.1 | ||||
| FLUX.1 [dev] | GPT4.1 | 73.7 | |||
| GPT4.1 | 73.7 | ||||
| Qwen2.5-VL | 70.6 | ||||
| FLUX.1 [dev] | Qwen2.5-VL | 68.5 | |||
| GPT4.1, CoT | 65.2 | ||||
| GPT4.1 | 65.1 | ||||
| Qwen2.5-VL | 62.1 | ||||
| Qwen2.5-VL, CoT | 62.0 | ||||
| Qwen2.5-VL | 62.0 | ||||
| GPT4.1 | 60.7 |
2 models have held number one on the official leaderboard since its release in September 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official PRISM-Bench (image generation) leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official PRISM-Bench (image generation) leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Vision
- MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy.
Qwen 3.6 Plus 86%
- MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
- CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
- MathVista In use Math reasoning in visual contexts (charts, geometry, puzzles) from 31 datasets. Answer accuracy.
Seed2.1 Pro 90.7%
- MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy.
Qwen 3.8 Max 97.7%
- MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy.
Qwen 3.8 Max 85.9%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 14 months ago, in September 2025, for BAGEL. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (18 entries).
- Pass Models scored 8 models we track, plus 10 other entries on its leaderboard.
- Warning Newest score The newest came 14 months ago, in September 2025, for BAGEL.
- Note Frontier models None of its models are on the capabilities index.
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on flux-reason-6m.github.io: 18 rows for models we track and 10 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.