The developing and developed nanofluid Rayleigh–Bénard flows between two parallel plates was simulated using the mesoscopic thermal lattice-Boltzmann method (LBM). The coupled effects of the thermal conductivity and the dynamic viscosity on the evolution of Rayleigh–Bénard flows were examined using different particle volume fractions (1–4%), while the individual effects of the thermal conductivity and the dynamic viscosity were tested using various particle sizes (11 nm, 20 nm, and 30 nm) and nanoparticle types (Al_{2}O_{3}, Cu, and CuO_{2}). Two different heating modes were also considered. The results show that Rayleigh–Bénard cell in nanofluids is significantly different from that in pure fluids. The stable convection cells in nanofluids come from the expansion and shedding of an initial vortex pair, while the flow begins suddenly in pure water when the Rayleigh number reaches a critical value. Therefore, the average Nusselt number increases gradually for nanofluids but sharply for pure liquids. Uniform fully developed flow cells with fewer but larger vortex pairs are generated with the bottom heating with nanofluids than with pure liquid, with extremely tiny vortexes confined near the top heating plate for top heating. The number of vortex pairs decreases with increasing nanoparticle volume fraction and particle diameter due to the increasing of dynamic viscosity. The average Nusselt number increases with the increasing Rayleigh number, while decreases with the increasing nanoparticle diameters. The nanoparticle types have little effect on the Rayleigh–Bénard flow patterns. The Rayleigh–Bénard flows are more sensitive with the dynamic viscosity than the thermal conductivity of nanofluids.