Medical image analysis has been transformed by deep learning, especially in dermatology, to allow AI-based computer vision models to classify and identify skin diseases with great accuracy. Notwithstanding its effectiveness, detection and diagnosis of Mpox remain challenging due to the lack of availability of datasets and visual likeness between Mpox and other dermatological conditions. To remedy this, we introduce a sophisticated deep learning-based Mpox detection system that utilizes both a new hybrid architecture and state-of-the-art Convolutional Neural Networks. Our work compares numerous prominent models, such as ResNet50 (97.25%), EfficientNetB3 (96.97%), Xception (96.15%), VGG16 (90.10%), and MobileNetV2 (89.83%). Moreover, we present a new hybrid ensemble model, Hybrid ResEffNet (ResNet50 + EfficientNetB3), which has an exceptional accuracy of 99.74%. This hybrid model surpasses standalone Convolutional Neural Networks, emphasizing the superiority of combining different architectures to achieve better performance. The models are trained on the upgraded dataset, Mpox Skin Lesion Dataset Version 2.0 (MSLD v2.0), comprising varied Mpox and non-Mpox skin lesion images. We utilized data augmentation and transfer learning methods to enhance model generalizability and resilience. Experimental findings confirm the efficiency of our strategy, with the suggested Hybrid ResEffNet model exhibiting better performance and accuracy than traditional Convolutional Neural Networks frameworks. In summary, our deep learning-based system provides a highly accurate and scalable Mpox detection solution. By combining deep learning with augmented data and a hybrid ensemble strategy, our approach supports early, consistent diagnosis, aiding healthcare workers in expedited disease screening and enhancing patient outcomes.