As the concept of CO2capture and storage has gained traction, some recent efforts have been made to build an accurate model for CO2viscosity prediction, which is an important element in CO2pipeline design and metering calculations. The present study focuses on the application of eleven computer-based models. There are a few important differences between this study and others that have utilized machine learning to determine CO2viscosity. This study is the most comprehensive investigation of this issue to date, using eleven intelligent models; unlike most of the literature models, the density of CO2is not used as an input parameter in the models; the best model's results are compared with the most recent experimentally derived correlations; it was shown that machine learning approaches not previously employed in studies outperform the model previously used in the literature; and instead of employing classic evolutionary approaches, two newly discovered optimization techniques (i.e., BAT and GOA) were applied. The performance of decision trees and random forest techniques is better than that of previously introduced best models for CO2viscosity prediction, such as multi-layer perceptron, radial basis function, and adaptive neuro-fuzzy inference systems, according to the results of this paper, which rank decision trees and random forest algorithms as the first and second-best algorithms among eleven methodologies studied. The final result of the decision trees method outperformed a current and most updated correlation for CO2viscosity prediction so far, as seen in various charts.