This paper introduces a novel deep learning architecture that combines the Fourier Neural Operator (FNO), BitNet quantization, ResNet, and Model-Agnostic Meta-Learning (MAML), along with embedded task representations to enhance image classification on CPUs. Unlike traditional models that rely on GPU acceleration, our proposed model leverages the spectral learning capabilities of FNO and the INT8 quantization of BitNet to deliver high performance while minimizing computational demands. On the CIFAR-10 dataset, the model achieves 99% training accuracy and 71.55% test accuracy, surpassing existing models compatible with CPU usage. These results highlight the model’s efficiency and adaptability, demonstrating strong potential for applications in edge computing. Additionally, the full implementation is open-source, promoting reproducibility and further research in this area. Keywords: Deep learning, Fourier Neural Operator, BitNet, quantization, image classification, meta-learning, CPU optimization.