Tip leakage loss significantly influences aerodynamic performance of high-pressure turbines, with squealer tips serving as an effective control strategy. The development of a tip leakage loss model is crucial for evaluating and predicting turbine aerodynamic performance and guiding blade tip design. This study presents a novel leakage loss model for squealer tips, employing a hybrid approach that integrates physics-driven and data-driven methodologies, followed by comprehensive validation. The leakage flow within the squealer tip gap is modeled into three basic flows: Vena contracta motion in both the pressure-side and suction-side squealer tip gaps, and jet diffusion inside the cavity. The specific flow pattern and loss magnitude inside the cavity are intricately linked to the evolution of the scraping vortex. The paper introduces a methodology for modeling the scraping vortex through deep learning, grounded in the separation and reattachment theory of backstep flow. Subsequently, it presents a physical model of the leakage flow across the squealer tip gap, informed by the classical theories of three identified basic flows and the scraping vortex's behavior. The influences of complex factors that are challenging to address solely through physical modeling are also taken into consideration with the aid of machine learning. The proposed model enables a rapid and precise prediction of key flow features, such as scraping vortex characteristics, discharge coefficient, leakage flow rate and momentum, alongside total leakage flow rate and leakage loss. This model provides a reliable analytical tool for predicting leakage performance and guiding designs for the squealer tip.