10K new test images following ImageNet collection process. Tests model generalization beyond the original test set.
Top 1 Accuracy is the reported evaluation metric for ImageNet-V2. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.
Higher is better
| Rank | Model | Trust | Score | Year | Links | Edit |
|---|---|---|---|---|---|---|
| 01 | Swin Transformer V2 Large | unverified | 84 | 2025 | Source ↗ | Edit result |
| 02 | swin-v2-large | paper | 84 | 2025 | Source ↗ | Edit result |
| 03 | ConvNeXt V2 Huge | unverified | 80.5 | 2025 | Source ↗ | Edit result |
| 04 | convnext-v2-huge | paper | 80.5 | 2025 | Source ↗ | Edit result |
Accuracy is the reported evaluation metric for ImageNet-V2. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.
Higher is better
| Rank | Model | Trust | Score | Year | Links | Edit |
|---|---|---|---|---|---|---|
| 01 | DINOv3 (7B) | unverified | 81.4 | 2025 | Paper ↗Code ↗ | Edit result |
| 02 | SigLIP 2 (g/16) | unverified | 79.8 | 2025 | Paper ↗Code ↗ | Edit result |
| 03 | ALIGN | unverified | 70.1 | 2021 | Paper ↗Code ↗ | Edit result |
| 04 | AltCLIP | unverified | 68.2 | 2022 | Paper ↗Code ↗ | Edit result |