A first step in assessing the impact of TAO data on operational models is to determine what percentage of the available data is actually being used in the models. The atmospheric models at NOAA's National Centers for Environmental Prediction (NCEP) have well-defined observation time and data receipt time cut-off windows within which the data must arrive to be assimilated. Analysis of the wind observations from the TAO array indicates that only 65% of the data set available to the models is being assimilated into the global atmospheric model suite at NCEP. The time of the observation, the on-off cycle times of the buoys, and the time between when the satellite is visible to the buoy and when it is visible to the ground station are the primary controlling factors of that percentage. The low percentage suggests that the method for real-time relay of data from the TAO array may not be optimum for operational applications. In this presentation the suite of global atmospheric models at NCEP, their data acceptance windows and how they differ among the models are described briefly. The quantity of TAO wind observations actually used in the models is shown for the period July/August 1996 and is compared to the total number of observations collected by the array during that same period. Illustrations are provided showing the number of observations used in the models displayed against several variables, including the time of the observation, the difference between the time of the observation and its time of arrival at NCEP, and the cycle times of the model. Recommendations are made on ways to increase the percentage of TAO observations used in the NCEP models.