Precipitation-use efficiency (PUE) is an important indicator used to determine how net primary productivity (NPP) responds to variation in precipitation, especially in arid and semi-arid ecosystems. The objective of this study was to determine the spatio-temporal patterns of PUE and its response to climatic factors along the precipitation gradient. The Carnegie-Ames-Stanford Approach (CASA) model was employed to simulate NPP in Inner Mongolia during 2001-2010 based on MOD13A1 data and spatially interpolated meteorological data. PUE was calculated as the ratio of NPP to annual precipitation. The results showed that:(1) The multi-year average PUE of grassland in Northwestern China was 0.68 g C m^-2 mm^-1. Among various type of temperate grassland, meadow steppe had the highest PUE, whereas the lowest PUE was found for desert. Furthermore, there were significant differences between the PUE of different grassland types. For the alpine grassland, alpine steppe had a higher PUE than alpine meadow. (2) Spatially, the PUE of temperate grassland increased first, peaking at~472.9 mm/a, and then decreased with the precipitation gradient (R²=0.65, P<0.001). The relationship between desert PUE and precipitation followed a similar trend (R²=0.63, P<0.001), in which the highest PUE was found in the regions with annual precipitation of 263.2 mm. For the alpine meadow, PUE was low at both the dry (>100 mm) and the wet ends of the annual precipitation gradient, and peaked around 559.2 mm (R²=0.47, P<0.001). (3) Temporally, the inter-annual variation of PUE also responded differently to climatic factors in different precipitation ranges. In the area with precipitation of 200-1000 mm, PUE was positively correlated with precipitation. For the regions where precipitation was higher than 1050 mm, temperature had much greater effects than precipitation on the inter-annual variations of PUE.