Coinjecting CO2 and light hydrocarbons with steam into oil sand reservoirs can improve the efficiency of the SAGD (steam assisted gravity drainage) process by reducing the steam oil ratio (SOR). The effects of these solvents on bitumen recovery enhancement depend on reservoir properties and operating conditions. To investigate the effects of solvents on bitumen viscosity in a solvent aided process, phase behaviors and viscosities of CO2-, C-3-, and C-4-bitumen systems were measured and modeled at high temperatures. Using the calibrated Peng Robinson equation of state (PR-EOS), the solubilities of solvents in the Clearwater bitumen sample from the Cold Lake region were predicted. High-pressure and high-temperature equipment using an electromagnetic-based viscometer was customized to measure the viscosities of CO2-, C-3-, and C-4-bitumen mixtures. The measured viscosity data were used to calibrate a nonlinear viscosity model which was used to predict liquid phase viscosity as a function of solvent solubility and temperature. The effects of solvent dissolution on bitumen viscosity were investigated using PR-EOS and the calibrated viscosity model. The results show that dissolving CO2, C-3, and C-4 in bitumen decreases its viscosity. This viscosity reduction is lowest and highest in the case of CO2 and C-4 dissolution, respectively. The effect of solvent dissolution on viscosity reduction is more pronounced at lower temperatures. EOS predictions and viscosity measurements indicate that increasing concentration of CO2, C-3, and C-4 above a certain threshold has a limited effect on reducing bitumen viscosity. At threshold solvent concentrations, bitumen viscosity can be reduced by 1.7, 5.6, and 15.2 times using CO2, C-3, and C-4, respectively, at 120 degrees C. Solubility and viscosity data suggest that C-4 has the potential to be used in hot-solvent recovery methods in shallow and deep oil sand reservoirs. C-3 may be a more effective solvent in deeper reservoirs which allow higher operating pressures. The modified viscosity model showed better performance than the Lobe and Shu correlations and logarithmic mixing rule. This model improves existing correlations for predicting viscosities of light solvent bitumen mixtures since it requires less input data and does not require density data.