ConceptEdit-Bench
Models edit images according to 1,000 instructions across six categories. A vision-language judge scores instruction following, visual quality, and consistency in unedited regions from 0 to 10. The per-case score averages those three dimensions.
The benchmark has 979 source images and a separate evaluation package. ConceptEdit-12M is the training dataset. The released scorer uses detailed English instructions by default.
- Made by
- Shanghai Jiao Tong University and Ant Group
- Released
- August 2026
- Kind
- Image generation and editing
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Vision, Image, video and 3D generation
- Last model added
- February 2026
- Scores refreshed
- November 18th, 2026
In the authors’ words
Existing image editing frameworks predominantly follow the training paradigm of text-to-image diffusion models. However, extending this paradigm to image editing highlights two inherent discrepancies, specifically, the insufficient attention to edit concept granularity and the training inefficiency caused by sparse supervision signals. To address these issues, we establish a comprehensive hierarchical taxonomy featuring over 1,000 fine-grained edit concepts and build ConceptEdit-12M, a massive dataset of 12 million high-quality editing pairs via an improved synthesis framework. This library-driven approach effectively rectifies the distribution collapse of generated data while ensuring high data fidelity. Furthermore, we propose a dense supervision training strategy that synthesizes multiple non-interfering concepts into single image pairs. By providing richer learning signals, this strategy significantly enhances both training efficiency and overall model performance. Training results validate our strategy, significantly outperforming prior works. Finally, we present ConceptEdit-Bench, a granular evaluation suite designed to diagnose model capabilities across a vast array of real-world scenarios.
Gemini 3.1 Flash Image leads with 66.2 among models released by February 2026 Gemini 3.1 Flash Image leads with 66.2 among models released by February 2026
Best score per entry from the benchmark's paper, read on November 18th, 2026. A paper’s table doesn’t add newer models. 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. This view shows only models we track, each at its best.
The newest model with a score here came out in February 2026, 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
- 1
Gemini 3.1 Flash Image 66.2
- 2 FireRed-Image-Edit-1.0 65.9
- 3
Seedream 4.5 63.3
- 4
Qwen-Image-Edit (2511) 63.3
- 5
Qwen-Image-Edit-2509 61.3
- 6 LongCat-Image-Edit 59.0
- 7 FLUX.1 Kontext [dev] 43.5
- 8
BAGEL-7B-MoT 41.7
Source: The ConceptEdit-Bench paper
- 1
Gemini 3.1 Flash Image 66.2
- 2
Seedream 4.5 63.3
- 3
Qwen-Image-Edit (2511) 63.3
- 4
Qwen-Image-Edit-2509 61.3
- 5 FLUX.1 Kontext [dev] 43.5
- 6
BAGEL-7B-MoT 41.7
Source: The ConceptEdit-Bench paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 66.2 | Feb 2026 | Official leaderboard | |
| 2 | FireRed-Image-Edit-1.0 | 65.9 | Official leaderboard | ||
| 3 | | 63.3 | Official leaderboard | ||
| 4 | | 63.3 | Official leaderboard | ||
| 5 | | 61.3 | Sep 2025 | Official leaderboard | |
| 6 | LongCat-Image-Edit | 59.0 | Official leaderboard | ||
| 7 | FLUX.1 Kontext [dev] | 43.5 | Official leaderboard | ||
| 8 | | 41.7 | May 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 66.2 | Feb 2026 | Official leaderboard | |
| 2 | | 63.3 | Official leaderboard | ||
| 3 | | 63.3 | Official leaderboard | ||
| 4 | | 61.3 | Sep 2025 | Official leaderboard | |
| 5 | FLUX.1 Kontext [dev] | 43.5 | Official leaderboard | ||
| 6 | | 41.7 | May 2025 | Official leaderboard |
Rankings by part
We read 6 parts from the benchmark’s paper, each ranking models on one part of the test, such as Advanced, Object and Composition. Gemini 3.1 Flash Image leads 5 of them, and Qwen-Image-Edit (2511) leads the other one. Arrows show how far a model moved from its overall rank among the models we track.
Advanced
6 models we track have a score here.
- 1
Gemini 3.1 Flash Image 71.8
- 2
Seedream 4.5 69.1
- 3
Qwen-Image-Edit-2509 ↑1 67.4
- 4
Qwen-Image-Edit (2511) ↓1 66.8
- 5 FLUX.1 Kontext [dev] 50.7
- 6
BAGEL-7B-MoT 42.2
Object
6 models we track have a score here.
- 1
Gemini 3.1 Flash Image 73.9
- 2
Qwen-Image-Edit (2511) ↑1 68.5
- 3
Seedream 4.5 ↓1 68.0
- 4
Qwen-Image-Edit-2509 63.8
- 5
BAGEL-7B-MoT ↑1 46.1
- 6 FLUX.1 Kontext [dev] ↓1 42.9
Composition
6 models we track have a score here.
- 1
Gemini 3.1 Flash Image 64.2
- 2
Seedream 4.5 63.6
- 3
Qwen-Image-Edit (2511) 60.4
- 4
Qwen-Image-Edit-2509 58.4
- 5 FLUX.1 Kontext [dev] 44.5
- 6
BAGEL-7B-MoT 38.9
Global
6 models we track have a score here.
- 1
Qwen-Image-Edit (2511) ↑2 70.4
- 2
Gemini 3.1 Flash Image ↓1 70.0
- 3
Seedream 4.5 ↓1 68.6
- 4
Qwen-Image-Edit-2509 64.6
- 5 FLUX.1 Kontext [dev] 49.0
- 6
BAGEL-7B-MoT 44.9
Portrait
6 models we track have a score here.
- 1
Gemini 3.1 Flash Image 56.2
- 2
Qwen-Image-Edit-2509 ↑2 54.3
- 3
Qwen-Image-Edit (2511) 53.8
- 4
Seedream 4.5 ↓2 52.9
- 5
BAGEL-7B-MoT ↑1 41.1
- 6 FLUX.1 Kontext [dev] ↓1 37.4
Text
6 models we track have a score here.
- 1
Gemini 3.1 Flash Image 75.3
- 2
Qwen-Image-Edit (2511) ↑1 73.5
- 3
Seedream 4.5 ↓1 71.3
- 4
Qwen-Image-Edit-2509 69.8
- 5 FLUX.1 Kontext [dev] 40.1
- 6
BAGEL-7B-MoT 33.9
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The ConceptEdit-Bench paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official ConceptEdit-Bench 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 9 months ago, in February 2026, for Gemini 3.1 Flash Image. 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 benchmark’s paper (8 entries).
- Pass Models scored 6 models we track, plus 2 other entries on its leaderboard.
- Warning Newest score The newest came 9 months ago, in February 2026, for Gemini 3.1 Flash Image.
- 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 arxiv.org: 6 rows for models we track and 2 other entries at the last read.
- Pass Last read November 18th, 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.