The viscosity of crude oils and their blends is a key parameter for studying hydrocarbon flow in reservoirs with gas injection and well analyzing the performance during gas lift. Even though several models have been developed to predict the viscosity of live crude oils, no study investigated the viscosity modeling of crude oils under gas injection. In this study, the one-parameter friction theory framework is combined with three characterization methods that were previously explored to model the viscosity of live oils (Khemka et al., Fuel, 2021, 283, 118926), and their predictive capabilities are further investigated for modeling the viscosity of blends of crude oil with five different gases. The performance of the recently developed SARA-based characterization using the Peng-Robinson (PR) equation of state (EoS) is compared with the SARA-based method using the Perturbed-Chain Statistical Association Fluid Theory (PC-SAFT) EoS and the single carbon number (SCN) method using the PR EoS by testing against 392 experimental viscosity data points from 23 different blends. The models' strong predictive capabilities are demonstrated by the fact that while only live oil data at saturation is required to fit simulation parameters, each model predicts the viscosity of various blends in the single-phase region with only 7% average error, which is satisfactory for practical applications. Even in the two-phase region, the predictions are within the experimental uncertainty; however, the SARA-based methods deliver slightly improved viscosity predictions even though they have a lower number of characterized components than the SCN method. Additionally, despite using the relatively simpler PR EoS with the SARA-based method, the viscosity predictions are at least as good as the predictions obtained using the highly advanced PC-SAFT EoS.