Deepfake technology, which uses artificial intelligence to create highly realistic synthetic media, poses significant threats to privacy, security, and the spread of misinformation. Traditional deepfake detection methods, primarily based on convolutional neural networks (CNNs), often fall short in effectively identifying these sophisticated forgeries. This project explores the use of Parallel Vision Transformers (PViTs) for deepfake detection, leveraging their advanced capabilities in modeling complex patterns and long-range dependencies in visual data understanding. We trained the PViT model on a dataset comprising of 140k real and fake faces using Google Colab with an Nvidia A100 GPU. Our results demonstrate that PViTs significantly enhance detection accuracy, precision, recall, and robustness, offering a promising solution for combating the challenges posed by deepfake technology attaining 91.92 accuracy.