Most weather and climate models are implemented in Fortran. Only a small handful of these models are actually able to leverage current and emerging high-performance computing infrastructures which often use graphics processing units (GPUs). In this talk we explore how implementing weather and climate models in high-level Python can be used to target a wide range of hardware architectures, deliver competitive performance and scalability, and enable exciting new workflows in research and development. Examples using two state-of-the-art global and regional weather and climate models (X-SHiELD and ICON) are given.