This study compares four surface wind products which may be used to force ocean circulation models of the North Pacific: wind stress derived from the Atlas special sensor microwave imager (SSM/I) based surface wind analyses (July 1987–June 1990), the Goddard Earth Observing System (GEOS) wind assimilation product at 10 m (March 1985–February 1990), and, at the appropriate overlapping times, the European Centre for Medium‐Range Weather Forecasts (ECMWF) wind analyses and the Comprehensive Ocean‐Atmosphere Data Set (COADS). Overall, there is good agreement in spatial and temporal variations between COADS and the model‐based analyses. The COADS product used in this study, based on monthly averaged winds, underestimates the monthly averaged stresses in the higher latitudes where the submonthly variance is high. In the tropics and subtropics where the variance is lower, the COADS stresses are closer to climatology than the weaker model‐based winds. The largest differences between the model‐based analyses occur during winter in the northernmost region of the North Pacific basin where the spatial and temporal variability is largest. The seasonal mean Atlas SSM/I stresses are stronger overall than the other model‐based analyses, with differences reaching as high as 0.09 N m−2in the high latitudes during winter. The Atlas product is also consistently stronger in the curl. Of the four curl products, the Atlas SSM/I and GEOS are closest in pattern and intensity in the tropics in both summer and winter. These also have the closest wavenumber spectral distribution in the high latitudes during winter. Integrations of a nonlinear, baroclinic quasi‐geostrophic ocean model using three of the wind stress curl distributions show differences of 30–45% in the mean integrated volume transport in the Kuroshio Extension. The Atlas SSM/I wind stress curl seems to force a more realistic subtropical gyre (in terms of transport and eddy energy levels) and western boundary current structure. The GEOS curl seems to force a more realistic subpolar gyre circulation. Overall, the assimilation of the SSM/I wind speed data improves the comparison between the model‐based analyses and ship‐based observations and produces a more realisti