Magnetic Resonance Imaging (MRI) is an important diagnostic tool for the diagnosis of brain tumors but is limited when achieving accurate classification and segmentation, especially with small sized or irregularly outlined tumors. To overcome these shortcomings, this study puts forward a hybrid model combining Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks with the objective of achieving maximum classification accuracy while maintaining minimal model simplicity. The suggested approach is compared with individual CNN, Random Forest (RF), Extreme Learning Machine (ELM), and LSTM models for comparison of performance. Experimental results demonstrate that the hybrid CNN–LSTM model outperforms the comparative approaches, with accuracy being 94.50%, precision being 94.55%, and recall being 94.51%. Moreover, the model provides strong ROC-AUC performance, good tumor localization, and real-time applicability. With a balance in terms of accuracy and computation efficiency, the proposed framework offers a strong diagnostic tool to support radiologists in the timely detection, improve treatment planning, and in turn, patient outcomes.