AUGMIX is a data processing method that improves robustness and uncertainty estimation in image classifiers. It is simple to implement, adds minimal computational overhead, and helps models withstand unforeseen corruptions. AUGMIX significantly improves robustness and uncertainty measures on challenging image classification benchmarks, closing the gap between previous methods and the best possible performance in some cases by more than half. The method combines stochasticity and diverse augmentations with a Jensen-Shannon Divergence consistency loss to achieve state-of-the-art performance. On CIFAR-10 and CIFAR-100, AUGMIX roughly halves the corruption robustness error of standard training procedures. On ImageNet, AUGMIX also achieves state-of-the-art corruption robustness and decreases perturbation instability. AUGMIX is easy to integrate into existing training pipelines and maintains or improves accuracy on standard benchmark datasets. The method uses a combination of augmentation chains, mixing, and a consistency loss to ensure robustness and uncertainty estimation. It is effective across various architectures and datasets, including CIFAR-10, CIFAR-100, ImageNet, and ImageNet-C. AUGMIX improves calibration and reduces flip probabilities, making it more reliable in real-world applications. The method is robust to data shifts and maintains performance even when the training and test distributions are mismatched. AUGMIX is a simple yet effective technique that enhances the robustness and uncertainty estimation of image classifiers.AUGMIX is a data processing method that improves robustness and uncertainty estimation in image classifiers. It is simple to implement, adds minimal computational overhead, and helps models withstand unforeseen corruptions. AUGMIX significantly improves robustness and uncertainty measures on challenging image classification benchmarks, closing the gap between previous methods and the best possible performance in some cases by more than half. The method combines stochasticity and diverse augmentations with a Jensen-Shannon Divergence consistency loss to achieve state-of-the-art performance. On CIFAR-10 and CIFAR-100, AUGMIX roughly halves the corruption robustness error of standard training procedures. On ImageNet, AUGMIX also achieves state-of-the-art corruption robustness and decreases perturbation instability. AUGMIX is easy to integrate into existing training pipelines and maintains or improves accuracy on standard benchmark datasets. The method uses a combination of augmentation chains, mixing, and a consistency loss to ensure robustness and uncertainty estimation. It is effective across various architectures and datasets, including CIFAR-10, CIFAR-100, ImageNet, and ImageNet-C. AUGMIX improves calibration and reduces flip probabilities, making it more reliable in real-world applications. The method is robust to data shifts and maintains performance even when the training and test distributions are mismatched. AUGMIX is a simple yet effective technique that enhances the robustness and uncertainty estimation of image classifiers.