Breast cancer is the most prevalent malignancy affecting women globally that poses a significant health burden, contributing to cancer-related mortality. Even though early detection and treatment strategies have improved over the years, Metastatic Breast Cancer (MBC) still remains one of the difficult cancers to treat. This is primarily due to the complexity of prediction and understanding its progression. Traditional prognostic models often depend on single-modal data, such as clinical features, and imaging scans. However, these approaches often fail to capture the complex biological and clinical landscape of the disease. Recent technological advancements have enabled the collection of diverse multi-modal data that can provide a comprehensive molecular and phenotypic characterization of breast tumors. This study leverages the power of deep learning, specifically a Convolutional Neural Network combined with Generative Adversarial Network (GAN) and Transfer Learning, applied to multi-modal data from The Cancer Genome Atlas BRCA (TCGA-BRCA) multi-omics dataset to enhance metastatic breast cancer prediction. Our model achieved an accuracy of 97.4%, demonstrating its potential for improved prognostication. Furthermore, a comparative analysis of state-of-the-art approaches from the past decade (2013–2024) was conducted to contextualize our findings and highlight the advancements offered by our multimodal deep learning approach.