Abstract: The comprehensive treatment of diseases through drug combinations is widely recognized as effective. However, current methods for discovering drug combinations primarily rely on screening drug target profiles and cell lines under dose-dependent conditions, which can be time-consuming, costly, and resource-intensive. Deep learning models have been introduced to predict the synergistic effects of drug combinations, offering improvements in therapeutic outcomes and reductions in side effects. These models have gradually become a key tool in predicting drug combinations and their efficacy. This article reviews deep learning-based approaches for predicting synergistic drug combinations, highlighting the strengths and limitations of various models and evaluating relevant performance metrics. These advanced methods not only demonstrate the cutting-edge potential of artificial intelligence in drug discovery but also offer new insights into traditional Chinese medicine, suggesting that modern technological approaches could unravel the complexities of traditional medicine combinations and support its modernization.