PoinTramba is a hybrid framework combining Transformer and Mamba for efficient point cloud analysis. The framework segments point clouds into groups, with the Transformer capturing intra-group dependencies and generating group embeddings. These embeddings are then processed by Mamba to capture inter-group relationships. A bi-directional importance-aware ordering (BIO) strategy reorders group embeddings based on their importance scores, enhancing Mamba's performance. The framework achieves a balance between computational efficiency and analytical performance, outperforming existing methods on ScanObjectNN, ModelNet40, and ShapeNetPart datasets. Extensive experiments demonstrate its effectiveness, establishing a new state-of-the-art benchmark in point cloud analysis. The method leverages the strengths of both Transformer and Mamba architectures, offering a new standard in the field. The code is available at https://github.com/xiaoyao3302/PoinTramba.PoinTramba is a hybrid framework combining Transformer and Mamba for efficient point cloud analysis. The framework segments point clouds into groups, with the Transformer capturing intra-group dependencies and generating group embeddings. These embeddings are then processed by Mamba to capture inter-group relationships. A bi-directional importance-aware ordering (BIO) strategy reorders group embeddings based on their importance scores, enhancing Mamba's performance. The framework achieves a balance between computational efficiency and analytical performance, outperforming existing methods on ScanObjectNN, ModelNet40, and ShapeNetPart datasets. Extensive experiments demonstrate its effectiveness, establishing a new state-of-the-art benchmark in point cloud analysis. The method leverages the strengths of both Transformer and Mamba architectures, offering a new standard in the field. The code is available at https://github.com/xiaoyao3302/PoinTramba.