Accurate prediction of pathological complete response (pCR) is useful for clinical precision treatment of lung cancer. Computed tomograph (CT) imaging is widely used for predicting pCR in lung cancer due to its rapid acquisition and ease of use. However, existing classification methods for pCR prediction are primarily limited to either convolutional neural networks (CNNs) or Transformer architectures, which can not capture effective global information or have relatively low computational efficiency. Therefore, this study proposes a novel CNN and Mamba hybrid network with multi-scale awareness module to achieve accurate pCR prediction on CT scans of lung cancer patients. In each stage of the proposed hybrid CNN-Mamba encoder, we first employ channel splitting to significantly reduce the number of parameters, and utilize the designed CNN branch and Mamba branch to extract effective local and global features. Specifically, in the Mamba branch, we propose a novel intra-slice and inter-slice scanning mechanism to implement 8-way 3D scanning to replace the original 2D scanning mechanism, thereby substantially enhancing the model’s ability to capture global information in 3D images. Furthermore, to better fuse CNN and Mamba features, we design a novel multiscale awareness feature fusion module with channel-level and multi-scale spatial level fusion. The proposed method is evaluated on a private dataset that includes CT scans of 108 lung cancer patients who undergone neoadjuvant chemoimmunotherapy. Experimental results demonstrate that our proposed HCM-Net achieves the best accuracy of 83.33% and AUC of 84.08%, outperforming other state-of-the-art methods.