Previously published models for predicting heavy oil viscosity based on the composition or API gravity are extensively evaluated to ascertain their effectiveness with 28 newly collected Kuwaiti heavy oil samples from different locations. The composition of each sample is determined, and its viscosity is measured at atmospheric pressure across a wide range of temperatures (20–80 °C), resulting in 196 data points. The measured viscosities are used to evaluate existing models based on the principle of corresponding states, the residual viscosity concept, and an equation of state (EOS). As all of them failed to accurately predict the Kuwaiti heavy oils being studied, a new compositional model is developed to overcome their inherent limitations. This new model provides a good prediction of the viscosity of Kuwaiti heavy oils, with an average relative error of 3.8%. In addition, it is much simpler and easier to use than existing methods when applied to Kuwaiti crudes.