Thermal runaway of chemical reactions often lead to secondary decomposition of the material, especially with autocatalytic decomposition characteristics. Autocatalytic decompositions are often characterized by drastic decomposition, intense heat release, high thermal risk and difficultly in control. Quantitative classification of different autocatalytic strengths is significant for the formulation of risk control measures and recognition of the most dangerous scenarios in relief system. In this work, the factor, which is the ratio of the heat released by the initial stage to that of the total reaction, has been defined to effectively and quantitatively assess the strength of autocatalytic decomposition based on the Perez-Benito model that includes two-stage competitive reactions. According to “3σ principle” of the normal distribution, the factor was classified into five grades: 0–5%, 5–18%, 18–82% 82–95%, 95–100% corresponding to the autocatalytic strength: strong, less strong, moderate, less weak and weak, respectively. This method is validated using cumyl hydro-peroxide and dicumyl peroxide, the autocatalytic strength of which are well known to be less strong (7.92%) and weak (100%), respectively. The results are consistent with the profile width, heat-quench-reheat test and the isothermal tests. This method was used to classify the autocatalytic strength of dimethyl sulfoxide, 29.5 mass% nitro-guanidine aqueous solution, and Benz aldoxime. The results show that the autocatalytic strengths of these three materials are strong (0.01%), moderate (31.52%) and less weak (82.67%), respectively. Through the classification assessment, it helps to get a more intuitive understanding of the degree of substance danger from the perspective of kinetics and thermodynamics.