Deep learning has revolutionized many domains, such as natural language processing, image recognition, and healthcare analysis, by providing accurate predictions and improved decision-making. However, choosing the best model architecture for a problem and selecting the problem domains is still a challenge. Convolutional neural networks (CNNs), long-short term memory networks (LSTMs), and dense neural networks are the models that have been used to build a conventional multi-model architecture for several problems. This study proposes a multimodel technique to determine the optimal model architecture based on the problem statement and input data type. The model selection process determines which model among the available models in the architecture is suitable for usage or in a combination of models. To assess the effectiveness of different models, we employ a diverse dataset comprising image data, text data, and numerical data. The proposed approach outperformed the existing state-of-the-art methods and obtained a maximum accuracy of 89% in image classification, which improved to 92% after our proposed ensemble technique, 90% in text sentiment analysis, and 88% in diabetes prediction with the standard datasets. The performance of the model was evaluated based on accuracy, sensitivity, selectivity, and training time. The proposed system helps in selecting the best for any type of input data samples.