• Benchmarks with strong baselines are a key ingredient for rapid progress on a problem. • Here, we define a benchmark for data-driven global, medium-range weather prediction. • The data is processed for convenient use in machine learning models and a quickstart guide is provided.

It is shown that it is possible to emulate the dynamics of a simple general circulation model with a deep neural network. After being trained on the model, the network can predict the complete model state several time steps ahead—which conceptually is making weather forecasts in the model world. Additionally, after being initialized with an arbitrary model state, the network can through repeatedly feeding back its predictions into its inputs create a climate run, which has similar climate statistics to the climate of the general circulation model. This network climate run shows no long‐term drift, even though no conservation properties were explicitly designed into the network.

Abstract. Recently, there has been growing interest in the possibility of using neural networks for both weather forecasting and the generation of climate datasets. We use a bottom–up approach for assessing whether it should, in principle, be possible to do this. We use the relatively simple general circulation models (GCMs) PUMA and PLASIM as a simplified reality on which we train deep neural networks, which we then use for predicting the model weather at lead times of a few days. We specifically assess how the complexity of the climate model affects the neural network's forecast skill and how dependent the skill is on the length of the provided training period. Additionally, we show that using the neural networks to reproduce the climate of general circulation models including a seasonal cycle remains challenging – in contrast to earlier promising results on a model without seasonal cycle.