Aboveground Biomass (AGB) of grassland is a key indicator reflecting the function and quality of grassland ecosystems. Estimating grassland AGB on a large scale is crucial for the management of grassland ecosystems. In this study, MODIS images were used as the data source to extract three different types of characteristic variables, such as reflectance, vegetation index and vegetation products. Combined with the field measured grassland AGB data, the parametric models represented by Multiple Linear Stepwise (MLR) regression and the nonparametric models which were Random Forest (RF), Support Vector Machine (SVM) and k Nearest Neighbor (kNN) models were constructed respectively to estimate and map the AGB of grassland in the Tibet Autonomous Region. The results showed that:(1) The multiple linear stepwise regression, random forest, support vector machine model and kNN reduced Root Mean Squared Error (RMSE) by 15.8%, 13.5%, 4.1% and 17.3% respectively after the inclusion of vegetation products feature variables, which showed that vegetation products as modeling variables for grassland AGB estimation could effectively improve the accuracy of estimation model. (2) Among the grassland AGB estimation models constructed from three combinations of variables, the model constructed from the combination of variables consisting of reflectance, vegetation index, and vegetation product worked best, with the kNN model having the highest estimation accuracy with R² of 0.60, RMSE of 0.43 t/hm², and Mean Absolute Error (MAE) of 0.34 t/hm². (3) The spatial distribution of grassland AGB showed the variation characteristics of lower in the northwest region, higher in the central region and more fragmented distribution pattern and higher in the eastern region. (4) The predicted values of MODIS vegetation products combined with the kNN model were basically consistent with the observed spatial distribution trend of AGB in grassland. In conclusion, MODIS vegetation products combined with the kNN model can be used as an effective reference for large scale grassland AGB remote sensing estimation.