Preserving human health and life is of utmost importance in the development of automatic detection methods for early brain tumor diagnosis, considering the severe neurological impairments and potential fatality associated with the disease. Computational efficiency plays a critical role in brain tumor classification for real-time decision-making, treatment planning, and overall healthcare system optimization. While convolutional neural networks (CNNs) are widely used for brain tumor detection due to their exceptional accuracy, their high computational demands present significant challenges. To address the challenge at hand, a hybrid model is employed, integrating a pre-trained convolutional neural network (CNN) transfer learning model and the Distributed computing programming paradigm. The primary objective involves two stages: In the first stage, Inception v3 and VGG19 CNN transfer learning models are deployed on GPUs for detecting brain malignancies. Performance metrics, including accuracy, precision, recall, and F1-Score, are assessed, along with a comparative analysis of computational time on CPUs and GPUs. Results show Inception v3 achieving a higher accuracy rate (approximately 98.83%) than VGG19 (77.65%), with superior computational speed on both CPU and GPU platforms. GPU execution significantly reduces computational time by up to 90%, attributed to the efficient architecture of Inception v3. In the second stage, real-time classification is conducted using Distributed computing process with previously trained CNN models for gliomas, meningiomas, and pituitary tumors, respectively. This integrated approach offers an efficient solution for real-time classification of large-scale brain tumor datasets.