The paper discusses the performance of different machine learning models and a deep learning model in forecasting annual Gross Domestic Product (GDP) per capita (PPP) data of 33 OECD countries using past year variables. It focuses on creating a universal forecasting model. For the analysis, the paper uses cross-country panel data consisting of 262 time-series variables with annual periodicity, including various growth, development, health, energy, finance, and social indicators and their lag terms for five years. The paper shows that the Artificial Deep Neural Network performed the best among the considered machine learning and deep learning models, followed by Gradient Boosted Regressor, whereas Ridge Regressor performed the worst. This paper gives insight into the application of machine learning and deep learning in forecasting GDP per capita. It shows how the further improvement of these computational methods and data availability would improve the forecast accuracy and precision.