Glaucoma is a leading cause of irreversible blindness worldwide, particularly among diabetic patients who are at higher risk due to systemic complications affecting ocular health. Early and accurate detection is essential to prevent vision loss; however, traditional diagnostic methods are time-consuming and often require expert interpretation. In this study, a novel Hybrid Multimodal Attention-based Glaucoma Prediction Model (HMAGPM) is proposed to improve prediction accuracy by integrating both retinal fundus images and clinical data. The model utilizes deep learning techniques, specifically convolutional neural networks such as EfficientNet or ResNet, to extract discriminative image features including optic disc structure, texture, and color variations. Simultaneously, clinical parameters such as HbA1c levels, blood pressure, age, and duration of diabetes are processed using machine learning algorithms. An attention mechanism is incorporated to focus on critical regions of the retina, particularly the optic nerve head, enhancing the model’s ability to detect glaucoma-related patterns. The extracted image and clinical features are combined through a feature fusion layer to form a comprehensive representation, which is then classified using a fully connected neural network. Additionally, explainable AI techniques such as Grad-CAM and SHAP are employed to provide visual and feature-level interpretations of the model’s predictions. Experimental results demonstrate that the proposed approach achieves high accuracy and reliability, outperforming conventional single-modal methods. The proposed system offers a robust, interpretable, and efficient solution for early glaucoma detection in diabetic patients, supporting clinical decision-making and reducing the risk of vision loss.