Dynamic route optimization is a critical challenge for the logistics and transportation industries, as companies strive to reduce operational costs, fuel consumption, and carbon emissions. Traditional route optimization methods, such as linear programming and heuristics, lack the flexibility to adjust to real-time changes in traffic conditions, vehicle type, and load balancing. This paper addresses the problem by employing a machine learning-based approach to dynamically optimize delivery routes using real-time GPS data, vehicle characteristics, and transportation distances. The proposed model uses a Gradient Boosting Regressor, trained on a dataset that includes variables such as origin and destination latitude/longitude, vehicle type, current location, and transportation distance. The model predicts the optimal routes based on these inputs, focusing on minimizing transportation distance as a proxy for fuel consumption and efficiency. Additionally, we introduced a classification task for predicting delivery delays, using binary classification to identify whether a delivery will be on time or delayed. The results show that the model achieves an RMSE of 199.47 km for predicting transportation distances, with an overall classification accuracy of 91% in predicting delivery delays. Feature importance analysis indicates that vehicle type and GPS coordinates are the most significant factors in determining route efficiency. The model’s performance implies that logistics companies can effectively reduce operational inefficiencies by adopting machine learning techniques for dynamic route optimization, which would lead to significant cost savings and environmental benefits.