The paper introduces 3D-Spectral-Spatial Mamba (3DSS-Mamba), a novel framework for hyperspectral image (HSI) classification. The 3DSS-Mamba leverages the State Space Model (SSM) to integrate long-range sequence modeling and linear computational efficiency, addressing the limitations of traditional Convolutional Neural Networks (CNNs) and Transformers in HSI classification. The framework consists of a Spectral-Spatial Token Generation (SSTG) module and multiple stacked 3D-Spectral-Spatial Mamba Blocks (3DMB). The SSTG converts HSI cubes into 3D spectral-spatial tokens, while the 3DSS mechanism performs pixel-wise selective scanning on these tokens to capture global spectral-spatial relationships. The 3DMB combines the 3DSS mechanism with conventional mapping operations to extract comprehensive spectral-spatial semantic features. Experimental results on three public HSI datasets (Pavia University, Indian Pines, and Houston 2013) demonstrate that the proposed method outperforms state-of-the-art methods in terms of classification accuracy, computational efficiency, and robustness. The 3DSS-Mamba is shown to be effective in capturing global spectral-spatial contextual dependencies, making it a promising solution for HSI classification tasks.The paper introduces 3D-Spectral-Spatial Mamba (3DSS-Mamba), a novel framework for hyperspectral image (HSI) classification. The 3DSS-Mamba leverages the State Space Model (SSM) to integrate long-range sequence modeling and linear computational efficiency, addressing the limitations of traditional Convolutional Neural Networks (CNNs) and Transformers in HSI classification. The framework consists of a Spectral-Spatial Token Generation (SSTG) module and multiple stacked 3D-Spectral-Spatial Mamba Blocks (3DMB). The SSTG converts HSI cubes into 3D spectral-spatial tokens, while the 3DSS mechanism performs pixel-wise selective scanning on these tokens to capture global spectral-spatial relationships. The 3DMB combines the 3DSS mechanism with conventional mapping operations to extract comprehensive spectral-spatial semantic features. Experimental results on three public HSI datasets (Pavia University, Indian Pines, and Houston 2013) demonstrate that the proposed method outperforms state-of-the-art methods in terms of classification accuracy, computational efficiency, and robustness. The 3DSS-Mamba is shown to be effective in capturing global spectral-spatial contextual dependencies, making it a promising solution for HSI classification tasks.