Abstract: The cone penetration test (CPT) is commonly used to evaluate sand liquefaction potential due to its efficiency and reliability. The typical CPT method (Chinese code method) shows unreasonable results in fine-grained and deep soils, and the results are difficult to be integrated into probabilistic risk assessments. In this study, a novel CPT hyperbolic probabilistic model is proposed to improve the Chinese code method based on the Bayesian weighted maximum likelihood estimation by considering the effects of fines content, the uncertainties and sampling bias. The critical cone tip resistance corresponding to a liquefaction probability (P_L) of 15% is recommended as the deterministic method. Case studies demonstrate the reliability of the proposed method. The results show that the critical cone tip resistance of the CPT hyperbolic model increases with the depth, and gradually stabilizes, which can better reflect the dynamic propertices of soils. The model can be applied to deep and high-fines-content soils, and the performance is better than the Chinese code method and widely-used Robertson and Wride's method. For sites with high uncertainties, the probabilistic models can be used as an alternative or supplement to deterministic ones.