One of the most prevalent and potentially deadly diseases affecting women worldwide is breast carcinoma. Prompt and accurate identification is essential for reducing mortality and improving treatment outcomes. This study presents a hybrid Convolutional Neural Network (CNN) model that is successful at diagnosing breast carcinoma from mammography images. The proposed model employs a feature fusion technique to integrate multiple pre-trained CNN architectures, using their complementing strengths. Through the extraction of both high-level and low-level spatial data, this hybridization improves classification speed. Transfer learning and fine-tuning techniques are used to adapt the model to two examples of medical imaging datasets: the DDSM mammography dataset and the MIAS dataset. Experimental evaluations show that the proposed hybrid model outperforms traditional single CNN models according to the precision of classification, precision, sensitivity, and F1-score. The model effectively reduces false positives, which is important in clinical applications where diagnosis accuracy is critical. A comparison with the traditional CNN architecture illustrates the hybrid approach’s robustness and adaptability. The proposed approach offers a reliable and scalable computer-aided diagnosis (CAD) tool by automating feature extraction and categorization, which aids radiologists in more accurate and efficient mammography interpretation.