Recognizing Bangla handwritten and printed characters poses unique challenges due to the complexity and diversity of Bangla script. This research aims to evaluate the performance of various deep learning-based models for Bangla character recognition, comparing Inception V3, VGG16, ResNet50 and MobileNet V2. The study uses a combined dataset of printed characters from Google and handwritten characters from the Ekush dataset. To enhance model performance, several data pre-processing and augmentation techniques were applied. Among the models tested, MobileNet V2 delivered the best performance, achieving a training accuracy of 98% and a test accuracy of 97.03%. This highlights the model’s potential for high-precision character recognition, making it well-suited for applications requiring both handwritten and printed Bangla text identification. The study concludes that deep learning approaches, particularly MobileNet V2, provide effective solutions for addressing the complexities of Bangla character recognition, enabling advancements in document digitization and automated language processing.