Molecular diffusion determines the time to reach local equilibrium in a reservoir. It can be a main production mechanism in scenarios such as production from fractured reservoirs or tight formation. However, there is a lack of high-pressure diffusion coefficients for reservoir fluids and its related systems. Many correlations exist, but there is no consensus on their accuracy for these systems. We provide a systematic review of the available data for systems related to reservoir fluids, as well as a comprehensive comparison of five commonly used correlations for hydrocarbon mixtures, including the extended Sigmund, Riazi-Whitson, Leahy-Dios-Firoozabadi, Wilke–Chang, and the Hayduk–Minhas correlations. We collected extensive data of diffusion coefficients in binary mixtures related to petroleum fluids and established a database of over 80 binaries and 1600 data points. We also collected the data for gas diffusion in different oils and reservoir fluids, but the data in high-pressure live oils are extremely scarce. The five correlations were evaluated using the binary database, and a few selected correlations using the oil database. None of the correlations show consistent and dominant superiority for all the binary mixtures, although some are better for particular groups/regions. For oils and reservoir fluids, the composition information is often incomplete. Only a few sets allow a comparison between different correlations. Although some trends can be identified from the correlation evaluation, no conclusive recommendation is made for a particular model, due to the data scarcity. The findings underscore the need for more accurate measurement and modeling of gas diffusion in mixtures that are more representative of reservoir fluids at high pressures.