A Vision Transformer base model with patch size 16, pretrained on ImageNet-21k with AugReg v2 and fine-tuned on ImageNet-1k for classification.