Large-scale machine learning models, for example FourCastNet, Pangu-Weather, GraphCast and ECMWF’s AIFS, are currently transforming weather forecasting and are also reshaping the weather and climate research landscape. The first machine learning models that can be applied to climate change scenarios are also being developed. This session will bring together developers and users from research and operations, from academia as well as private enterprises, to discuss the current state of the art and future developments in the field. We welcome contributions from machine learning model developers, consortia and users of single component machine learning models, such as those focused on the atmosphere, ocean or sea ice, as well as coupled models that consider the entire Earth system. Foundation models, which learn more general representations of the Earth system, and studies on the combination of large-scale machine learning models and traditional physics-based solvers are also welcome. Of particular interest are also contributions on verification methods, out-of-distribution experiments, real or idealized case studies across different scales (e.g. air pollution, solar energy production, extratropical cyclones, ocean dynamics) as well as contributions with a focus on the physical consistency of such machine learning models. The session aims to facilitate dialogue between researchers interested in scientific discovery and developers interested in novel machine learning ideas pertinent to the domain, e.g. spatio-temporal diffusion models, variational autoencoders and novel training protocols.