The quality assessment of protein models is a key technology in protein structure prediction and has become a prominent research focus in the field of structural bioinformatics since advent of CASP7. Model quality assessment method not only guides the refinement of protein structure model but also plays a crucial role in selecting the best model from multiple candidate conformations, offering significant value in biological research and practical applications. This study begins with reviewing the critical assessment of protein structure prediction (CASP) and continuous automated model evaluation (CAMEO), and model evaluation metrics for monomeric and complex proteins. It primarily summarizes the development of model quality assessment methods in the last five years, including consensus methods (multi-model methods), single-model methods, and quasi-single-model methods, and also introduces the evaluation methods for protein complex models in CASP15. Given the remarkable progress of deep learning in protein prediction, the article focuses on the in-depth application of deep learning in single-model methods, including data set generation, protein feature extraction, and network architecture construction. Additionally, it presents the recent efforts of our research group in the field of model quality assessment. Finally, the article analyzes the limitations and challenges of current protein model quality assessment technology, and also looks forward to future development trends.