This document introduces Fashion-MNIST, a new dataset created by Zalando Research as a drop-in replacement for the MNIST dataset for benchmarking machine learning algorithms. Fashion-MNIST consists of 60,000 training images and 10,000 test images of 10 fashion product categories, formatted similarly to MNIST. It was created to address issues with MNIST being too easy and not representative of modern computer vision tasks. The dataset has gained popularity in the machine learning community as an alternative to MNIST, with over 2,000 stars on GitHub and being supported by several machine learning libraries.