Understanding the thermophysical properties for mixtures of CO2 and hydrocarbons at reservoir conditions is very important for the correct design and optimization of CO2-enhanced oil recovery and carbon storage in depleted oil or gas fields. In this paper, we present a comprehensive thermodynamic study of the prototype system (CO2 + n-heptane) comprising highly-accurate measurements of the saturated-phase densities, compressed-fluid densities, and bubble and dew points at temperatures from 283 K to 473 K and pressures up to 68 MPa over the full range of composition. We use these results to examine the predictive capability of two leading thermodynamic models: the Predictive Peng-Robinson (PPR-78) equation of state and a version of the Statistical Associating Fluid Theory for potentials of the Mie form, known as SAFT-γ Mie. Both of these models use group contribution approaches to estimate interaction parameters and can be applied to complex multi-component systems. The comparison shows that both approaches are reliable for the phase behavior. Neither model is entirely satisfactory for density, with each exhibiting absolute average relative deviations (AARD) from the experimental data of about 4% for the saturated-phase densities and 2% for the compressed-fluid densities; however, SAFT-γ Mie is found to be much more accurate than PPR-78 for the compressibility, with an overall AARD of 6% compared with 18% for PPR-78.