Breast cancer continues to be a highly common and serious health issue for women globally, highlighting the critical demand for precise and non-invasive diagnostic methods. Thermal imaging provides a non-invasive and economical method for the early identification of abnormal heat patterns linked to malignant tissues. This study presents a streamlined Convolutional Neural Network (CNN) model tailored for the classification of thermal breast images into cancerous and non-cancerous categories. The model underwent training and evaluation using the publicly accessible DMR-IR thermal image dataset, showcasing exceptional classification performance with an accuracy of 99%, and achieving precision, recall, and F1-scores of 0.99, complemented by flawless AUC and average precision scores. The design is refined for maximum efficiency, rendering it ideal for use in environments with limited resources, like rural clinics or integrated diagnostic systems. The experimental findings confirm the model’s robustness and scalability, indicating significant promise for real-time breast cancer screening. Our efforts connect accessible thermal imaging with automated diagnosis driven by deep learning, enhancing early detection rates and minimizing diagnostic burdens.