This dataset accompanies the article titled "Model-based deep reinforcement learning for accelerated learning from flow simulations". The archives contained in this dataset are: - data.tar.gz: the main research results (trajectory data, snapshots of policy and value networks, log files of the training process, results of repeated independent runs (seeds)) - mbdrl-main-branch-Apr2024.zip: a snapshot of the code repository used to create all derived results presented in the article; the repository is also available at https://github.com/JanisGeise/MB_DRL_for_accelerated_learning_from_CFD - drlfoam-fork-mbdrl-Apr2024.zip: drlFoam is the library implementing the DRL infrastructure and logic; this archive contains a snapshot of the drlFoam fork used in this work; the fork adds model-based learning to drlFoam; the fork is also available at https://github.com/JanisGeise/drlfoam/tree/mb_drl