<p>P91 steel is widely used due to its excellent properties such as oxidation resistance and high-temperature steam corrosion resistance. However, center cracks may occur during the solidification process. The solute equilibrium partition coefficient, a key input parameter in solidification, not only affects solidification micro-segregation calculations but also determines the accuracy of hot cracking susceptibility evaluation. In this study, based on the solidification micro-segregation model, focusing on the solute equilibrium partition coefficient, using the brittle temperature range (<i>BTR</i> criterion) and |<i>dT/d(f</i>_s^0.5)| (<i>Kou</i> criterion) as hot cracking susceptibility indices, a variable solute equilibrium partition coefficient model suitable for 91 steel is established. The variation law of equilibrium partition coefficients of each component is analyzed by FactSage software during solidification, the relationship between each component’s equilibrium partition coefficient and interface temperature is fitted, and the influence of component equilibrium partition coefficients on micro-segregation prediction and hot cracking susceptibility evaluation are examined. The results show that for the problems of low accuracy and poor applicability when micro-segregation models evaluate hot cracking susceptibility, the variable equilibrium partition coefficient model fully considering inter-component interactions highly matches literature values. On the basis of optimizing liquidus correlation, the variable equilibrium partition coefficient model further improves the accuracy of P91 steel hot cracking susceptibility evaluation. The deviation between <i>BTR</i> calculated and experimental values optimizes from 1.52 to 0.87&#xa0;pct; the deviation between <i>Kou</i> calculated and experimental values improves from −&#xa0;1.94 to 1.03&#xa0;pct.</p>