The paper introduces Interpretable3D, an ad-hoc interpretable classifier designed for 3D point cloud data. It addresses the lack of explainability in 3D models, which is crucial for decision-critical applications. Unlike existing *post-hoc* explanations that can be misleading, Interpretable3D provides reliable and intuitive explanations by selecting the most similar prototype for new samples. The method involves two iterative training steps: Prototype Estimation, where prototypes are updated with the mean of embeddings within the same sub-class, and Prototype Optimization, where prototypes are penalized or rewarded based on their performance. The prototypes are further updated with the most similar observations in the final epochs, enhancing their interpretability. Interpretable3D is evaluated on four popular point cloud models (DGCNN, PointNet2, PointMLP, and PointNeXt) and shows comparable or superior performance compared to softmax-based models in 3D shape classification and part segmentation tasks. The code is available at: github.com/FengZicai/Interpretable3D.The paper introduces Interpretable3D, an ad-hoc interpretable classifier designed for 3D point cloud data. It addresses the lack of explainability in 3D models, which is crucial for decision-critical applications. Unlike existing *post-hoc* explanations that can be misleading, Interpretable3D provides reliable and intuitive explanations by selecting the most similar prototype for new samples. The method involves two iterative training steps: Prototype Estimation, where prototypes are updated with the mean of embeddings within the same sub-class, and Prototype Optimization, where prototypes are penalized or rewarded based on their performance. The prototypes are further updated with the most similar observations in the final epochs, enhancing their interpretability. Interpretable3D is evaluated on four popular point cloud models (DGCNN, PointNet2, PointMLP, and PointNeXt) and shows comparable or superior performance compared to softmax-based models in 3D shape classification and part segmentation tasks. The code is available at: github.com/FengZicai/Interpretable3D.