This study presents the development of an improved technique for viscosity measurements under high pressure. The apparatus is based on the principle of the falling ball viscometer, implemented in a high-pressure autoclave fitted with visualisation windows. The originality here is that the balls fall through a tube open at both ends with a diameter slightly greater than that of the balls, allowing a simplified modelling and numerical simulation. A numerical approach has been used for viscosity determination. Calculations have been made with COMSOL Multiphysics^® with the laminar Navier–Stokes model for Newtonian mixtures. It includes the specific hydrodynamic effects without the need for a calibration fluid. However, validation experiments were carried out at atmospheric pressure with dimethylsulfoxide (DMSO) at 298, 308 and 318 K and with cocoa butter at 313 and 353 K, with values of viscosity in the range from 1.4 to 45.4 mPa s. Comparative measurements with literature data have been conducted with cocoa butter saturated with carbon dioxide at 313 and 353 K and for pressures ranging from 0.1 to 25 MPa. At 313 K, viscosity varies from 45.4 mPa s to 3.1 mPa s while at 353 K it varies from 12.4 to 1.9 mPa s. For both isotherms tested, within the range 0–15 MPa, the higher the CO₂ dissolution in the cocoa butter, the lower the viscosity. However, this decrease in viscosity is more pronounced at the lowest temperature. Above 15 MPa the CO₂ dissolution effect on viscosity becomes insignificant, i.e. within the experimental error, due to a counter effect linked with the high hydrostatic pressure. Furthermore, the limits of use of this method have been determined. This technique is revealed as reliable and can therefore be used with other binary systems.