Compared to other characteristics of CO2 exchange, gross primary productivity (P g ) is most directly related to photosynthetic activity. Until recently, it was considered difficult to obtain measurement-based P g . The objective of our study was to evaluate if P g can be estimated from continuous CO2 flux measurements using nonlinear identification of the nonrectangular hyperbolic model of ecosystem-scale, light-response curves. Estimates of P g and ecosystem respiration (R e ) were obtained using Bowen ratio– energy-balance measurements of CO2 exchange in a true-steppe ecosystem in northern Kazakhstan during four growing seasons (1998–2001). The maximum mean weekly apparent quantum yield (αmax) was 0.0388 mol CO2 mol photons and the maximum mean weekly P g was 28 g CO2/m2/day in July 2000. The highest mean weekly R e max (20 g CO2m2/day) was observed in July of both 1999 and 2000. Nighttime respiration calculated from daily respiration corrected for length of the dark period and temperature (using Q 10 = 2) was closely associated with measured nighttime respiration (R 2 = 0.67 to 0.93). The 4-year average annual gross primary production (GPP) was 1617 g CO2/m2/ year (range = 1308–1957). Ten-day normalized difference vegetation index corrected for the start of the season (NDVIsos) was closely associated with 10-day average P g (R 2 = 0.66 to 0.83), which was higher than R 2 values for regressions of mean 10-day net daytime fluxes on NDVIsos (0.55–0.72). This demonstrates the advantage of usingP g in scaling up flux-tower measurements compared to other characteristics (net daytime flux or net 24-h flux).