Optimum fertilizer-use decision requires supporting tool such as crop-nutrient response functions. The objectives of this research in Ethiopia were to develop tef nutrient response functions for selected AEZ, assess variability in nutrient response functions among the AEZ, and predict and evaluate the goodness-of-fit between observed and calculated tef grain yield for N and P nutrient rates. Geo-referenced tef fertilizer trials data of 147 and 139 trials for N and P, respectively, documented by OFRA project from past and recent trial results were used for analysis. Trials conducted on Vertisols and Cambisols were considered. The independent variables that influence the dependent variable (yield) were elevation, location, soil properties, climate, and their square or two-way interactions. Data were subjected to Generalized Linear Model at p ≤ 0.05. The asymptotic crop-nutrient response function was adopted. The nutrient response functions were determined for each AEZ and so used to predict tef yield response to N and P nutrient rates. The goodness-of-fit between predicted and observed yield data were evaluated using R2, RMSE, and slope of fitting lines. The highest and lowest coefficients of tef nutrient response functions were reported for AEZ6 and AEZ3, respectively. The paired mean differences comparisons of the coefficients of tef crop-nutrient response functions among some AEZ were shown significant differences. The yield predicted using tef crop-nutrient response functions with respect to N and P nutrient rates ranged from 0.55 to 18 ton ha−1 which is highly significant difference. The goodness-of-fit between the predicted yield using nutrient response functions and observed yield showed R2 > 0.91, RMSE < 0.2 and slope near 1. Such values indicate that the nutrient response functions predictive ability is acceptable for optimum nutrient rates decision-making in the study area conditions. The tef crop-nutrient response functions (predictive equations) can thus be used to estimate tef yield response to the nutrient rates instead of conducting field trials.