The paper introduces PyFormer, a novel approach that combines the strengths of Pyramid and Vision Transformer for Hyperspectral Image Classification (HSIC). PyFormer addresses the challenges of variable-length input sequences in traditional Transformer models, which are common in HSIC. By organizing input data into hierarchical segments, each representing different abstraction levels, PyFormer enhances processing efficiency for lengthy sequences. At each level, a dedicated transformer module captures both local and global context, facilitating communication and abstraction propagation within the hierarchy. Experimental results demonstrate that PyFormer outperforms traditional approaches, particularly on challenging datasets with limited training data. The method's robustness and generalizability make it promising for real-world applications in HSIC. The source code for PyFormer is available at https://github.com/mahmood00/PyFormer.The paper introduces PyFormer, a novel approach that combines the strengths of Pyramid and Vision Transformer for Hyperspectral Image Classification (HSIC). PyFormer addresses the challenges of variable-length input sequences in traditional Transformer models, which are common in HSIC. By organizing input data into hierarchical segments, each representing different abstraction levels, PyFormer enhances processing efficiency for lengthy sequences. At each level, a dedicated transformer module captures both local and global context, facilitating communication and abstraction propagation within the hierarchy. Experimental results demonstrate that PyFormer outperforms traditional approaches, particularly on challenging datasets with limited training data. The method's robustness and generalizability make it promising for real-world applications in HSIC. The source code for PyFormer is available at https://github.com/mahmood00/PyFormer.