60K 32x32 color images in 10 classes. Classic small-scale image classification benchmark with 50K training and 10K test images.
Accuracy is the reported evaluation metric for CIFAR-10. 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
Muted rows were not state of the art when published — an earlier or same-year result already scored better.
| Rank | Model | Trust | Score | Year | Links | Fix |
|---|---|---|---|---|---|---|
| 01 | ViT-H/14 (JFT-300M) | verified | 99.5 | 2026 | Source ↗ | Looks wrong? |
| 02 | ViT-L/16 (JFT-300M) | verified | 99.42 | 2026 | Source ↗ | Looks wrong? |
| 03 | BiT-L (ResNet152x4) | verified | 99.37 | 2026 | Source ↗ | Looks wrong? |
| 04 | ViT-H/14 (IN-21K) | verified | 99.27 | 2026 | Source ↗ | Looks wrong? |
| 05 | deit-b-distilled | vendor | 99.1 | 2026 | Source ↗ | Looks wrong? |
| 06 | ViT-L/16 (IN-21K) | verified | 99 | 2026 | Source ↗ | Looks wrong? |
| 07 | EfficientNet-B8 (NoisyStudent) | verified | 98.7 | 2026 | Source ↗ | Looks wrong? |
| 08 | convnext-v2-base | vendor | 98.7 | 2026 | Source ↗ | Looks wrong? |
| 09 | ViT-B/16 (IN-21K) | verified | 98.13 | 2026 | Source ↗ | Looks wrong? |
| 10 | Swin-B | verified | 98 | 2026 | Source ↗ | Looks wrong? |
| 11 | resnet-50 | vendor | 96.01 | 2026 | Source ↗ | Looks wrong? |