Early diagnosis of Alzheimer’s disease allows patients to receive early and effective treatment as a way to increase their chances of survival. We propose CA-ConvNeXt for Early Alzheimer’s disease classification to solve the common MCI, AD, and NC classification problems. We employ the latest ConvNeXt network, which has a simpler topology and greater performance than ResNet and Swin Transformer. We effectively increase the model performance and reach 96$$\%$$% accuracy on the public ADNI dataset by adding Coordinate Attention to the ConvNeXt network.