In this article we present a study comparing deep learning models. Multilayer Perceptron (MLP) Recurrent Neural Network (RNN) Bidirectional Long Short Term Memory (bi-LSTM) and Graph Neural Network (GNN). Within the realm of session-based recommender systems (SBRS). Our goal is to investigate how well these standard models perform in predicting user preferences based on interactions within sessions using real-time data. The assessment will focus on analyzing the strengths and weaknesses of each model in addressing the fleeting nature of user interactions commonly seen in SBRS. Our analysis reveals that bi-LSTM models generally exhibit the accuracy and reliability in capturing user preferences and session dynamics. Additionally, the simpler MLP model also demonstrates performance occasionally surpassing its complex counterparts. The findings shed light on the benefits and limitations of deep learning models for enhancing recommender systems, indicating promising avenues for future research and development in personalized content delivery systems.