VIF-Bench
Models generate images from multiple references and visual instructions. Judges check whether the result follows controls such as pose or lighting without copying unwanted features.
- Made by
- The University of Tokyo
- Released
- September 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
- September 2025
- Scores refreshed
- November 20th, 2026
GPT-Image-1.5 leads with 7.60 among models released by September 2025 GPT-Image-1.5 leads with 7.60 among models released by September 2025
Best score per entry from the benchmark's paper, read on November 20th, 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 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
- 1 GPT-Image-1.5 7.60
- 2 GPT-Image-1 7.55
- 3 Nano Banana Pro Multi-step 7.40
- 4
Gemini 2.5 Flash Image Preview (nano-banana) 7.13
- 5
Qwen-Image-Edit (2511) 4.94
- 6 DreamOmni2 4.64
- 7 FLUX.1 Kontext 4.61
- 8
Qwen-Image-Edit-2509 3.93
Source: The VIF-Bench paper
- 1 GPT-Image-1.5 7.60
- 2 GPT-Image-1 7.55
- 3
Gemini 2.5 Flash Image Preview (nano-banana) 7.13
- 4
Qwen-Image-Edit (2511) 4.94
- 5
Qwen-Image-Edit-2509 3.93
Source: The VIF-Bench paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-Image-1.5 | 7.60 | Official leaderboard | ||
| 2 | GPT-Image-1 | 7.55 | Official leaderboard | ||
| 3 | Nano Banana Pro | 7.40 | Multi-step | Official leaderboard | |
| 4 | | 7.13 | Official leaderboard | ||
| 5 | | 4.94 | Official leaderboard | ||
| 6 | DreamOmni2 | 4.64 | Official leaderboard | ||
| 7 | FLUX.1 Kontext | 4.61 | Official leaderboard | ||
| 8 | | 3.93 | Sep 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-Image-1.5 | 7.60 | Official leaderboard | ||
| 2 | GPT-Image-1 | 7.55 | Official leaderboard | ||
| 3 | | 7.13 | Official leaderboard | ||
| 4 | | 4.94 | Official leaderboard | ||
| 5 | | 3.93 | Sep 2025 | Official leaderboard |
Rankings by part
We read 12 parts from the benchmark’s paper, each ranking models on one part of the test, such as Text Instruction Following, Qwen3-VL: Text Instruction Following and Reference Consistency. GPT-Image-1.5 leads 7 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.
Text Instruction Following
5 models we track have a score here.
- 1 GPT-Image-1.5 6.79
- 2 GPT-Image-1 6.51
- 3
Gemini 2.5 Flash Image Preview (nano-banana) 6.40
- 4
Qwen-Image-Edit (2511) 3.01
- 5
Qwen-Image-Edit-2509 2.54
Qwen3-VL: Text Instruction Following
5 models we track have a score here.
- 1 GPT-Image-1.5 8.27
- 2 GPT-Image-1 8.01
- 3
Gemini 2.5 Flash Image Preview (nano-banana) 7.62
- 4
Qwen-Image-Edit (2511) 4.21
- 5
Qwen-Image-Edit-2509 3.21
Reference Consistency
5 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑2 8.39
- 2 GPT-Image-1.5 ↓1 8.12
- 3 GPT-Image-1 ↓1 7.84
- 4
Qwen-Image-Edit (2511) 3.38
- 5
Qwen-Image-Edit-2509 2.85
Qwen3-VL: Reference Consistency
5 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑2 9.48
- 2 GPT-Image-1.5 ↓1 9.34
- 3 GPT-Image-1 ↓1 9.27
- 4
Qwen-Image-Edit (2511) 4.93
- 5
Qwen-Image-Edit-2509 3.64
Visual Instruction Adherence
5 models we track have a score here.
- 1
Gemini 2.5 Flash Image Preview (nano-banana) ↑2 5.23
- 2 GPT-Image-1.5 ↓1 4.57
- 3 GPT-Image-1 ↓1 4.35
- 4
Qwen-Image-Edit (2511) 2.50
- 5
Qwen-Image-Edit-2509 2.31
Qwen3-VL: Visual Instruction Adherence
5 models we track have a score here.
- 1 GPT-Image-1.5 7.90
- 2
Gemini 2.5 Flash Image Preview (nano-banana) ↑1 7.76
- 3 GPT-Image-1 ↓1 7.40
- 4
Qwen-Image-Edit (2511) 3.02
- 5
Qwen-Image-Edit-2509 2.59
Visual Instruction Cleanliness
5 models we track have a score here.
- 1 GPT-Image-1.5 9.80
- 2 GPT-Image-1 9.69
- 3
Qwen-Image-Edit (2511) ↑1 7.78
- 4
Gemini 2.5 Flash Image Preview (nano-banana) ↓1 7.09
- 5
Qwen-Image-Edit-2509 6.89
Qwen3-VL: Visual Instruction Cleanliness
5 models we track have a score here.
- 1 GPT-Image-1.5 9.85
- 2 GPT-Image-1 9.81
- 3
Qwen-Image-Edit (2511) ↑1 7.65
- 4
Gemini 2.5 Flash Image Preview (nano-banana) ↓1 7.46
- 5
Qwen-Image-Edit-2509 6.76
Scene Coherence
5 models we track have a score here.
- 1 GPT-Image-1.5 8.24
- 2 GPT-Image-1 7.99
- 3
Gemini 2.5 Flash Image Preview (nano-banana) 7.13
- 4
Qwen-Image-Edit (2511) 5.62
- 5
Qwen-Image-Edit-2509 4.31
Qwen3-VL: Scene Coherence
5 models we track have a score here.
- 1 GPT-Image-1 ↑1 8.95
- 2 GPT-Image-1.5 ↓1 8.87
- 3
Gemini 2.5 Flash Image Preview (nano-banana) 7.93
- 4
Qwen-Image-Edit (2511) 5.08
- 5
Qwen-Image-Edit-2509 3.37
Visual Quality
5 models we track have a score here.
- 1 GPT-Image-1 ↑1 8.90
- 2 GPT-Image-1.5 ↓1 8.89
- 3
Gemini 2.5 Flash Image Preview (nano-banana) 8.55
- 4
Qwen-Image-Edit (2511) 7.33
- 5
Qwen-Image-Edit-2509 4.68
Qwen3-VL: Visual Quality
5 models we track have a score here.
- 1 GPT-Image-1.5 9.74
- 2 GPT-Image-1 9.72
- 3
Gemini 2.5 Flash Image Preview (nano-banana) 9.52
- 4
Qwen-Image-Edit (2511) 8.75
- 5
Qwen-Image-Edit-2509 5.54
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official VIF-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 14 months ago, in September 2025, for Qwen-Image-Edit-2509. 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 5 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 14 months ago, in September 2025, for Qwen-Image-Edit-2509.
- 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 3 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.