Grain size dependence of secondary creep rate in pure metals and single-phase alloys and the mechanisms/models to interpret the dependence have been reviewed. Two types of the grain size dependence were reported, both of which show a negative dependence approximately below 100 μm. The model proposed by Garofalo in 1960s assumes that the density of dislocations generated at grain boundaries and sub-boundaries determines the secondary creep rate, which is not experimentally supported. Mclean’s model considers preferential subgrain growth near grain boundaries, which might be important in practical steels and alloys with sub-grains such as high Cr ferritic heat resistant steels. Grain boundary sliding (GBS) and its accommodation process in grain is considered as a source of the negative grain size dependence. A finite element modeling performed by Crossman and Ashby, which simulated a deformation process where GBS is accommodated by the power law creep process in the grain interior, indicates that the negative grain size dependence cannot be interpreted by the accommodation process. Based on substructure observation and internal stress measurements, Terada established a “core and mantle” model where the mantle region near grain boundary has no internal stress. This model reasonably interprets the negative grain size dependence.