Estimates of pasture production are critical for comparing treatment effects in research trials, as inputs to decision support models, and for on-farm use in constructing grazing plans. We compared two techniques for measuring pasture production, over 2 years in 36 paddocks continuously grazed with sheep, on hill country at the Ballantrae Research Station. One technique (Cut method) involved harvesting pasture regrowth after trimming to a standard height. The other (Model method) calculated pasture production as the sum of grazing animal intake (estimated using a spreadsheet-based model with inputs including animal performance and pasture quality) and changes in pasture cover (above-ground biomass). Annual pasture production estimates were reasonably well correlated (R2 = 0.66), but those using the Cut method were 55% greater than using the Model method. Agreement was best in spring and summer, and poorest in autumn and winter. Factors contributing to these discrepancies included the contrasting starting conditions for the pasture at the start of each measurement period (shorttrimmed for the Cut method, and patch-grazed for the Model method); the effect of grazing in the Model method maintaining pastures in a more vegetative state; and the greater losses from senescence and decay in the Model method pastures. The potential for errors using the Model method was high where measurement periods were short, as accuracy of measurement of pasture cover can be suspect. The Model method probably better quantified net pasture production in grazed pastures. Keywords: grazing, hill pasture, intake, model, pasture production