Image Classification2012en
ImageNet Large Scale Visual Recognition Challenge 2012
1.28M training images, 50K validation images across 1,000 object classes. The standard benchmark for image classification since 2012.
Current State of the Art
CoCa (finetuned)
91
top-1-accuracy
top-1-accuracy Progress Over Time
Showing 5 breakthroughs from Dec 2015 to May 2022
Key Milestones
May 2022
CoCa (finetuned)Current SOTA
Current SOTA on ImageNet-1K. 2.1B parameters. Contrastive Captioner architecture.
91.0
+0.6%
Total Improvement
15.8%
Time Span
6y 6m
Breakthroughs
5
Current SOTA
91.0
Top Models Performance Comparison
Top 10 models ranked by top-1-accuracy
Best Score
91.0
Top Model
CoCa (finetuned)
Models Compared
10
Score Range
8.0
top-1-accuracyPrimary
| # | Model | Score | Paper / Code | Date |
|---|---|---|---|---|
| 1 | CoCa (finetuned)Open Source Google | 91 | Dec 2025 | |
| 2 | ViT-G/14Open Source Google | 90.45 | Dec 2025 | |
| 3 | ConvNeXt V2 HugeOpen Source Meta | 88.9 | Dec 2025 | |
| 4 | ViT-H/14Open Source Google | 88.55 | Dec 2025 | |
| 5 | Swin Transformer LargeOpen Source Microsoft | 87.3 | Dec 2025 | |
| 6 | EfficientNetV2-LOpen Source Google | 85.7 | Dec 2025 | |
| 7 | DeiT-B DistilledOpen Source Meta | 85.2 | Dec 2025 | |
| 8 | EfficientNet-B7Open Source Google | 84.4 | Dec 2025 | |
| 9 | DeiT-BOpen Source Meta | 83.1 | Dec 2025 | |
| 10 | ConvNeXt V2 TinyOpen Source Meta | 83 | Dec 2025 | |
| 11 | ViT-L/16Open Source Google | 82.7 | Dec 2025 | |
| 12 | ViT-B/16Open Source Google | 81.2 | Dec 2025 | |
| 13 | ResNet-50 (A3 training)Open Source Timm | 80.4 | Dec 2025 | |
| 14 | ResNet-152Open Source Microsoft | 78.6 | Dec 2025 | |
| 15 | EfficientNet-B0Open Source Google | 77.1 | Dec 2025 | |
| 16 | ResNet-50Open Source Microsoft | 76.15 | Dec 2025 |