To make up for the limitations and improve the accuracy of 1D nuclear-magnetic-resonance (NMR) logging in the evaluation of formation fluid properties, 2D NMR logging has become the focus of research. Increasing the sequence and inversion parameters of the 2D NMR can effectively improve the antinoise properties and resolution of the inversion, but at the same time, the reduced inversion speed and increased memory occupied will put forward higher requirements on the computer configuration and add to the cost of calculation, which poses challenges to the application of the traditional 2D NMR inversion algorithms. In view of the above defects, we have developed a new fast 2D NMR inversion LSQR-RSVD hybrid algorithm, and we have used the nonnegative least-squares (LSQR) calculation result as the initial value of the RSVD inversion. Taking oil-water and gas-water models as examples, the 2D NMR inversion effects of (T2, D), (T1, T2) are analyzed in detail, and (T1, D) is also discussed by several groups of echo trains with variable echo interval (TE) and waiting time (TW). Compared to the inversion algorithm commonly used, the new hybrid algorithm can improve the inversion speed and significantly reduce the memory occupancy. Its remarkable advantages can further promote the application of 2D and even multidimensional NMR logging in practice.