FlowMol is a flow matching model for 3D de novo molecule generation that achieves improved performance over prior flow matching methods and is competitive with state-of-the-art diffusion models while exhibiting a >10-fold reduction in inference time. The paper explores the use of flow matching, a generative modeling framework that generalizes diffusion models, for de novo molecule generation. Flow matching provides flexibility in model design but assumes continuously-valued data. However, 3D de novo molecule generation requires jointly sampling continuous and categorical variables such as atom positions and atom types. To address this, the authors extend the flow matching framework to categorical data by constructing flows constrained to the probability simplex, called SimplexFlow. However, experiments show that a simpler approach that does not account for the categorical nature of the data yields equivalent or superior performance. FlowMol, a flow matching model for 3D de novo generative modeling, is presented, which achieves improved performance over existing flow matching methods and is competitive with diffusion models. The paper also raises important questions about the design of prior distributions for achieving strong performance in flow matching models. The code and trained models for reproducing this work are available at https://github.com/dunni3/FlowMol.FlowMol is a flow matching model for 3D de novo molecule generation that achieves improved performance over prior flow matching methods and is competitive with state-of-the-art diffusion models while exhibiting a >10-fold reduction in inference time. The paper explores the use of flow matching, a generative modeling framework that generalizes diffusion models, for de novo molecule generation. Flow matching provides flexibility in model design but assumes continuously-valued data. However, 3D de novo molecule generation requires jointly sampling continuous and categorical variables such as atom positions and atom types. To address this, the authors extend the flow matching framework to categorical data by constructing flows constrained to the probability simplex, called SimplexFlow. However, experiments show that a simpler approach that does not account for the categorical nature of the data yields equivalent or superior performance. FlowMol, a flow matching model for 3D de novo generative modeling, is presented, which achieves improved performance over existing flow matching methods and is competitive with diffusion models. The paper also raises important questions about the design of prior distributions for achieving strong performance in flow matching models. The code and trained models for reproducing this work are available at https://github.com/dunni3/FlowMol.