Benchmark Almanac

SuperCLUE-Image

ArtArk 95.1 tops its leaderboard GPT-Image-2.5 Sunburst 94.3 leads the models we track
Not in model cards

Models generate images from Chinese prompts. The benchmark scores text rendering, visual quality, prompt adherence and scene composition.

SuperCLUE publishes scores for 25 models in its 2026 results workbook.

Made by
SuperCLUE
Released
June 2024
Kind
Image generation and editing
Level
Everyday
Scored
Graded by a model
Answers
Open-ended
Measures
Capability
Language
Chinese
In and out
Text → Images
Interaction
Single turn
Run by
A host
Scores refreshed
November 21st, 2026
25
Entries on its own leaderboard, 10 models we track

ArtArk tops the leaderboard with 95.1 GPT-Image-2.5 Sunburst leads with 94.3

Best score per entry from the benchmark's own leaderboard, read on November 21st, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is GPT-Image-2.5 Sunburst, at 94.3. This view shows only models we track, each at its best.

  1. 1 ArtArk 95.1
  2. 2 GPT-Image-2.5 Sunburst 94.3
  3. 3 SenseNova U1 Pro 93.8
  4. 4 GPT-Image-2.5 Flare 93.5
  5. 5 GPT-Image-2 93.2
  6. 6 Nano Banana 2.1 92.9
  7. 7 Doubao-Seedream-5.0-pro 91.9
  8. 8 Qwen-Image-3.0-Pro 91.6
  9. 9 MAI-Image-2.6 91.2
  10. 10 Grok Imagine Image 2.0 90.3
  11. 11 SenseNova U1.5 Lite 90.2
  12. 12 Gemini-3.1-Flash-Image (Nano Banana 2) 90.1
  13. 13 JT4.1-Image 89.7
  14. 14 Doubao-Seedream-5.0-lite 89.4
  15. 15 Qwen-Image-2.0-Pro 86.7
  16. 16 Nano Banana Pro 84.9
  17. 17 Doubao-Seedream-4.5 84.7
  18. 18 Hy-Image-3.0 77.8
  19. 19 Recraft V4.1 Utility Pro 76.3
  20. 20 Wan2.6-t2i 75.2
  21. 21 Kling-v3 66.0
  22. 22 Kling-v2-1 63.1
  23. 23 FLUX.2 [max] 61.9
  24. 24 FLUX.2 [pro] 58.9
  25. 25 CogView-4 46.3

Source: Official SuperCLUE-Image leaderboard

#EntryScoreSettingDateSource
1 ArtArk 95.1 Official leaderboard
2 GPT-Image-2.5 Sunburst 94.3 Official leaderboard
3 SenseNova U1 Pro 93.8 Official leaderboard
4 GPT-Image-2.5 Flare 93.5 Official leaderboard
5 GPT-Image-2 93.2 Official leaderboard
6 Nano Banana 2.1 92.9 Official leaderboard
7 Doubao-Seedream-5.0-pro 91.9 Official leaderboard
8 Qwen-Image-3.0-Pro 91.6 Official leaderboard
9 MAI-Image-2.6 91.2 Official leaderboard
10 Grok Imagine Image 2.0 90.3 Official leaderboard
11 SenseNova U1.5 Lite 90.2 Official leaderboard
12 Gemini-3.1-Flash-Image (Nano Banana 2) 90.1 Official leaderboard
13 JT4.1-Image 89.7 Official leaderboard
14 Doubao-Seedream-5.0-lite 89.4 Official leaderboard
15 Qwen-Image-2.0-Pro 86.7 Official leaderboard
16 Nano Banana Pro 84.9 Official leaderboard
17 Doubao-Seedream-4.5 84.7 Official leaderboard
18 Hy-Image-3.0 77.8 Official leaderboard
19 Recraft V4.1 Utility Pro 76.3 Official leaderboard
20 Wan2.6-t2i 75.2 Official leaderboard
21 Kling-v3 66.0 Official leaderboard
22 Kling-v2-1 63.1 Official leaderboard
23 FLUX.2 [max] 61.9 Official leaderboard
24 FLUX.2 [pro] 58.9 Official leaderboard
25 CogView-4 46.3 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-Image-2.5 Sunburst 94.3 Official leaderboard
2 GPT-Image-2.5 Flare 93.5 Official leaderboard
3 GPT-Image-2 93.2 Official leaderboard
4 Qwen-Image-3.0-Pro 91.6 Official leaderboard
5 MAI-Image-2.6 91.2 Official leaderboard
6 Grok Imagine Image 2.0 90.3 Official leaderboard
7 Qwen-Image-2.0-Pro 86.7 Official leaderboard
8 Recraft V4.1 Utility Pro 76.3 Official leaderboard
9 FLUX.2 [max] 61.9 Official leaderboard
10 FLUX.2 [pro] 58.9 Official leaderboard

Rankings by part

We read 6 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as 图像质量, 图文一致性 and 汉字生成. GPT-Image-2.5 Sunburst leads 3 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

图像质量

10 models we track have a score here.

  1. 1 MAI-Image-2.6 ↑4 91.4
  2. 2 GPT-Image-2.5 Flare 90.8
  3. 3 GPT-Image-2.5 Sunburst ↓2 90.0
  4. 4 GPT-Image-2 ↓1 89.8
  5. 5 Grok Imagine Image 2.0 ↑1 87.4
  6. 6 Qwen-Image-2.0-Pro ↑1 87.0
  7. 7 Qwen-Image-3.0-Pro ↓3 86.6
  8. 8 FLUX.2 [max] ↑1 84.0
  9. 9 Recraft V4.1 Utility Pro ↓1 80.0
  10. 10 FLUX.2 [pro] 79.4

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

No dated values yet.

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

All 636 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

7 of 8 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark

Catalog record 1 warning 1 pass
  • Warning Task access We don’t know whether the tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the official leaderboard (25 entries).
  • Pass Models scored 10 models we track, plus 15 other entries on its leaderboard.
  • Note Newest score No scores yet.
  • 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 superclueai.com: 10 rows for models we track and 15 other entries at the last read.
  • Pass Last read November 21st, 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.

How we check