Brain tumor MRI data resource is considered for research and the application of different artificial neural network pre-trained architectures using the strategy of transfer learning is employed to understand the working and adaptability. The deep learning architectures are developed and tested in the jupyter notebook, Google Colab environment using python programming language and keras libraries – “Inception ResNet v2, ResNet 152 v2 and VGG19”. The feature extraction for developed models was done using the above architectures and finetuned for the classification phase into four possible outcomes of brain tumor (GL, MG, PT, NoTumor). Inception ResNet V2 achieved a promising accuracy in test phase of 92.21%, with a robust training accuracy of 98.18%. VGG19 and ResNet 152 V2 models followed closely with testing accuracy of 91.64%, and accuracies during training of 97.16% and 98.82%. The findings emphasize the capability of Inception ResNet V2 for high-precision tumor classification, with ResNet 152 V2 also showing strong training performance. Overall, the strength of transfer-learning in medical imaging, especially for brain tumor detection, where the challenge is to access vast labeled datasets is highlighted in this exploration. By providing accurate and reliable classification, these implementations can significantly support radiologists and clinicians in early diagnosis and personalized treatment planning.