The success of low field Nuclear Magnetic Resonance (LF-NMR) to estimate heavy oil properties depends on a good selection of mathematical models and fitting parameters. Since the correlations proposed are not universally applicable, in this study, a NMR published model was chosen and tuned to determine the density and viscosity of several mixtures of a Colombian heavy oil with toluene. The process began by mixing toluene with heavy oil to obtain several measuring points with properties similar to those of heavy oils. Each mixture was taken to a 7.5 MHz spectrometer at 40°C, where NMR parameters were acquired and used in the five pre-selected mathematical models. The reliability of viscosity measurements was analysed with the root mean square error (RMSE) and maximum absolute error (MAE). After the NLS regression process, the most accurate prediction was reached through the Burcaw model, with RMSE values of 40.55 cP. On the other hand, the density was estimated with the Wen correlation with results showing a relative error percentage of less than 1%. According to such error values, the tuned models are considered a starting point to extend the NRM technique use to other Colombian heavy oils with low uncertainty levels.