Groundwater is an often overlooked freshwater resource compared to surface water, but groundwater is used widely across the United States, especially during periods of drought. If groundwater models can successfully simulate past conditions, they may be used to evaluate potential future pumping scenarios or climate conditions, thus providing a valuable planning tool for water-resource managers. Quantifying the groundwater-use component for a groundwater model is a vital but often challenging endeavor. This dataset includes groundwater withdrawal rates modeled for the Ozark Plateaus aquifer system (Ozark system) from 1900 to 2010 by groundwater model cell (2.6 square kilometers) for five water-use divisions: agriculture (including irrigation and aquaculture), livestock, public supply (including municipal and rural water districts), and non-agriculture (including thermoelectric power generation, mining, commercial, and industrial). Domestic groundwater withdrawal rates are available in the complementary dataset “Domestic groundwater withdrawal rates from the Ozark Plateaus aquifer system, 1900 to 2010”. The Ozark system is located in the central United States and is composed of interbedded Cambrian to Pennsylvanian clastic and carbonate lithologies. In stratigraphic order, the Ozark system includes the Basement confining unit, St. Francois aquifer, St. Francois confining unit, Ozark aquifer, Ozark confining unit, Springfield Plateau aquifer, and Western Interior Plains confining system. Generally, the lower portion of the Ozark aquifer is the primary source of groundwater across much of Missouri and the Springfield Plateau aquifer is used across northern Arkansas. A full description of the methods used to model groundwater withdrawal rates from the Ozark system are available in Knierim et al. (2017) (see larger work citation). Briefly, groundwater use was modeled by 1) acquiring site-specific and county-level groundwater withdrawal rates and well locations (with and without pumping information) from state agencies and the U.S. Geological Survey, 2) linearly interpolating groundwater withdrawal rates to create a yearly time-step for the period of observations (generally 1962 to 2010), 3) extrapolating county-level groundwater withdrawal rates to 1900 for non-agriculture, agriculture, and livestock groundwater use by assuming a linear decrease from the oldest, recorded groundwater withdrawal rate (generally between 1962 and 1985) to 0 million liters per day (ML/d) in 1900, 4) extrapolating site-specific groundwater withdrawal rates to 1900 for public supply assuming use was linearly related to population change, then constraining groundwater withdrawal rate to 0 ML/d in 1900 using a multiplier that incrementally ranged from zero in 1900 to one in 2010, 5) attributing groundwater withdrawal rates to well locations using a hierarchical process where observed site-specific groundwater withdrawal rates were used first, followed by county-level groundwater withdrawal rates disaggregated to wells where pumping was known to occur at any time, and lastly county-level groundwater withdrawal rates disaggregated to well locations with a potential groundwater-use type based on land use, and 6) aggregation into model cells (row, column, layer) by summing modeled site-specific groundwater withdrawal rates using well location and depth. The large dataset (148,836 well locations) and long period (110 years) necessitated modeling groundwater use programmatically using Python 2.7.