The genotypic status of gliomas is a critical indicator for assessing patient prognosis and survival duration, with its genetic classification holding significant clinical value. Although biomolecular markers serve as essential criteria for glioma classification, challenges such as high detection costs and low efficiency persist. In recent years, the rapid advancement of artificial intelligence (AI) has provided novel solutions for the automated classification of gliomas, with MRI-based deep learning techniques for genetic status prediction emerging as a research hotspot. Current studies have established a technical framework ranging from supervised to unsupervised learning by integrating multimodal MRI data, achieving high predictive accuracy in single-gene classification through multi-task joint optimization and transfer learning. Meanwhile, multi-gene joint analysis, as an emerging direction, has yielded preliminary exploratory results. In the future, key breakthroughs in glioma genetic classification research will include the development of multi-gene synchronous prediction models, enhancing generalization capabilities for small samples through generative data augmentation, improving prediction accuracy using multi-source heterogeneous data and large models, and introducing visualization techniques to enhance interpretability.