VideoMamba is a state space model designed for efficient video understanding. It addresses the challenges of local redundancy and global dependencies in video understanding by adapting the Mamba model, which uses a selective state space model (SSM) to achieve linear complexity and efficient long-term modeling. VideoMamba outperforms existing 3D convolutional neural networks and video transformers, offering scalability, sensitivity for short-term action recognition, superiority in long-term video understanding, and compatibility with other modalities. It achieves these capabilities through a novel self-distillation technique, enhanced temporal sensitivity, and efficient processing of long videos. VideoMamba is evaluated on various datasets, including ImageNet-1K, Kinetics-400, and SthSthV2, demonstrating its effectiveness in both short-term and long-term video understanding. It also performs well in multi-modal tasks, such as video-text retrieval, showing robustness and adaptability. VideoMamba's linear complexity allows it to handle high-resolution long videos efficiently, making it a promising solution for video understanding. The model is open-sourced, providing a scalable and efficient solution for comprehensive video understanding.VideoMamba is a state space model designed for efficient video understanding. It addresses the challenges of local redundancy and global dependencies in video understanding by adapting the Mamba model, which uses a selective state space model (SSM) to achieve linear complexity and efficient long-term modeling. VideoMamba outperforms existing 3D convolutional neural networks and video transformers, offering scalability, sensitivity for short-term action recognition, superiority in long-term video understanding, and compatibility with other modalities. It achieves these capabilities through a novel self-distillation technique, enhanced temporal sensitivity, and efficient processing of long videos. VideoMamba is evaluated on various datasets, including ImageNet-1K, Kinetics-400, and SthSthV2, demonstrating its effectiveness in both short-term and long-term video understanding. It also performs well in multi-modal tasks, such as video-text retrieval, showing robustness and adaptability. VideoMamba's linear complexity allows it to handle high-resolution long videos efficiently, making it a promising solution for video understanding. The model is open-sourced, providing a scalable and efficient solution for comprehensive video understanding.