Gliomas are the most common malignant brain tumors driven by genetic and microenvironmental factors, with newly recognized high-risk molecular subtypes requiring aggressive treatment. This study aims to address the limitations of biopsy-based molecular typing by developing a non-invasive multimodal magnetic resonance imaging habitat imaging model to predict high-risk subtypes, thereby improving early detection and guiding treatment. Data of 204 glioma patients retrieved from The Cancer Genome Atlas public database were retrospectively analyzed. Habitat imaging based on K-means clustering was applied to three habitat regions in pre-operative T1CE and T2FLAIR sequences, extracting 10,416 radiomics features. Analysis of variance was used to assess the correlation between features and labels, screening radiomics features significantly associated with high-risk molecular subtypes. A support vector machine classifier was employed to construct a habitat radiomics model. Logistic regression (LR) was used to identify relevant clinical features, and a clinical prediction model was established, followed by performance evaluation. A combined model was developed by integrating the habitat radiomics model and the clinical model using multivariate LR. The predictive performance of the three models was evaluated and compared using metrics such as the area under the receiver operating characteristic curve (AUC), decision curve analysis (DCA), and calibration curves. The combined model achieved AUC = 0.943 and 0.912 in the training and test sets, respectively, outperforming the clinical model (training set: AUC = 0.830; test set: AUC = 0.841) and the habitat radiomics model (training set: AUC = 0.914; test set: AUC = 0.864). In DCA, the combined model demonstrated significantly higher and more stable net benefits within a reasonable clinical threshold range compared to the other two models. Calibration curves indicated that the combined model also exhibited superior calibration performance. This study shows that combining clinical and radiomics data improves glioma risk prediction, but multicenter validation of such approach for clinical use is warranted.