Abstract: Glutenite reservoir is characterized by large particle size range, low porosity and permeability, complex pore structure and strong heterogeneity, so it is difficult to predict its quality classification and productivity. Based on NMR logging, this paper proposes the classification scheme and productivity prediction method of glutenite reservoirs. Using the transverse relaxation time (T₂) distribution measured by NMR experiment of rocks, the paper extracts the characteristic parameters of T₂ distribution such as percentages of three-pore components (percentage of small pore components S₁, percentage of medium pore components S₂ and percentage of large pore components S₃) and geometric mean of T₂ distribution (T_{2_LM}), and builds the reservoir quality index (I_RQ) comprehensively reflecting the pore structure and physical properties of rock. Characteristic parameters of T₂ distribution and reservoir quality index have a good correlation with pore throat radius, displacement pressure and other mercury penetration experiment parameters, which can quantitatively characterize the rock pore structure. The classification chart of glutenite reservoirs has been established using S₃/S₁ and I_RQ. The productivity prediction comprehensive index (F) and the productivity prediction model of glutenite reservoir have been built based on S₃/S₁, T_{2_LM}, I_RQ and effective thickness (H) of the reservoir. The classification method and productivity prediction model of glutenite reservoirs based on NMR logging have achieved a good application effect in the glutenite reservoir of Baikouquan Formation in the west slope of Mahu sag, Junggar Basin, and the effectiveness and practicability of the method have also been verified, wichi is of certain reference significance for the research of glutenite reservoir classification and productivity.