Rock thermal conductivity (TC) is a critical parameter in geothermal energy exploration and geological engineering, and is strongly influenced by the mineralogical and geochemical composition of rocks. In this study, granite from the Huangshadong geothermal field in Huizhou, Guangdong Province, was selected as the primary research subject. Combined with supplementary data from other regions, this work systematically investigates the quantitative relationships between granite thermal conductivity and both its mineral and chemical compositions. Moreover, the performance and applicability of four commonly used hybrid models for predicting rock thermal conductivity were comprehensively evaluated. The results reveal a strong positive correlation between granite thermal conductivity and quartz content, and a significant negative correlation with biotite content. Chemically, thermal conductivity exhibits a notablely positive correlation with SiO₂ content, and negative correlations with Al₂O₃, CaO, MgO, and TiO₂ contents. These findings suggest that evaluating rock thermal conductivity from a geochemical perspective offers a more nuanced understanding than mineralogical analysis alone. Among the evaluated models, the Harmonic Mean model demonstrated the best predictive performance for granite thermal conductivity, with an average mean error (AME) of 9.06 and a root mean square error (RMSE) of 0.34. However, the coefficient of determination (R²) for all four models remained in the range of 0.43–0.44, indicating moderate predictive accuracy. This limitation underscores the significant influence of rock structural characteristics, which are not fully captured by composition-based models. Future research should focus on the enhancement of predictive accuracy of granite thermal conductivity models by integrating both compositional and structural parameters. This study provides valuable insights into the microscale mechanisms governing granite thermal conductivity, contributing to a deeper scientific foundation for geothermal resource assessment and development.