PointMamba is a novel state space model (SSM) designed for point cloud analysis, leveraging the success of the Mamba model from natural language processing. Unlike traditional Transformers, PointMamba employs a linear complexity algorithm, offering global modeling capabilities while significantly reducing computational costs. The method uses space-filling curves to effectively tokenize point clouds and adopts a simple, non-hierarchical Mamba encoder as its backbone. Comprehensive evaluations demonstrate that PointMamba outperforms various Transformer-based methods on multiple datasets, achieving superior performance while reducing GPU memory usage and FLOPs. This work highlights the potential of SSMs in 3D vision tasks and provides a simple yet effective baseline for future research. The code is available at <https://github.com/LMD0311/PointMamba>.PointMamba is a novel state space model (SSM) designed for point cloud analysis, leveraging the success of the Mamba model from natural language processing. Unlike traditional Transformers, PointMamba employs a linear complexity algorithm, offering global modeling capabilities while significantly reducing computational costs. The method uses space-filling curves to effectively tokenize point clouds and adopts a simple, non-hierarchical Mamba encoder as its backbone. Comprehensive evaluations demonstrate that PointMamba outperforms various Transformer-based methods on multiple datasets, achieving superior performance while reducing GPU memory usage and FLOPs. This work highlights the potential of SSMs in 3D vision tasks and provides a simple yet effective baseline for future research. The code is available at <https://github.com/LMD0311/PointMamba>.