Early and accurate detection is important in improving patient outcomes and a major global health issue today is breast cancer. This research introduces an innovative deep-learning approach to significantly enhance diagnostic accuracy in breast cancer detection using histopathological images. We address this challenge by developing a hybrid deep learning model that combines the strengths of ResNet-18 and ResNet- 34 architectures, to the best of our knowledge, this hybrid design hasn’t been used before for classifying breast cancer histopathological images. The methodology involved rigorous data preprocessing to ensure quality, followed by training on a 70–30 split, with optimization through Cyclic Learning Rate and L2 regularization or weight decay to enhance performance and prevent overfitting. This hybrid model achieved a remarkable 91.77% accuracy, 91.32% precision, 88.15% recall, 89.70% F1-Score, and 0.97 AUC, surpassing both individual ResNet models and demonstrating a substantial improvement over traditional diagnostic methods, as well as existing single deep learning approaches. The implementation of this technology promises to transform breast cancer diagnostics by enabling earlier intervention, potentially leading to significantly improved survival rates and reduced mortality, marking a major step forward in the application of AI in healthcare. This innovative strategy offers a scalable and efficient solution, holding the potential to benefit healthcare systems and patients worldwide.