This paper examines the ability of three ocean mixed layer submodels to depict inter-annual variations of sea surface temperature (SST) in a global configuration of the HYbrid Coordinate Ocean Model (HYCOM). The mixed layer submodels are (1) the K-Profile Parameterization (KPP), (2) the NASA Goddard Institute for Space Studies (GISS) turbulence closure, and (3) the Mellor-Yamada Level 2.5 (MY) turbulence closure. Accuracy of SSTs from the submodels is investigated during 1996-2001, which includes the onset of the strong 1998 La Niña event, when a record cold SST anomaly in the eastern equatorial Pacific occurred. The model simulations (with no ocean data assimilation or relaxation to SST climatology) reveal that all three submodels generally capture the westward extent of the SST cooling within the eastern equatorial Pacific during the transition period from the 1997 El Niño to the 1998 La Niña, one of the largest short term events ever observed (7°C change in SST from May to June 1998). During the six-month period after the transition, the daily SST from the submodels is ≈ 2°C warmer than the buoy SSTs obtained from the Tropical Atmosphere Ocean (TAO) array. Some of these biases are due to deficiencies in the net shortwave radiation and near-surface air temperature used for the simulations. Finally, comparisons with 166 yearlong daily SST time series from many buoys over various regions of the global ocean, including mostly equatorial Pacific, give median RMS differences of 0.65°, 0.70°, and 0.78°C for KPP, GISS, and MY, respectively, during 1996-2001.