<p>Since significant progress has been made in reducing other systemic errors, the multipath effect has become a bottleneck for improving global instantaneous or single-epoch multi-constellation and multi-frequency precise point positioning (PPP) cascading ambiguity resolution (AR) (PPP-CAR). We compared the residual variance reduction percentages of multi-constellation and multi-frequency multipath models constructed with sidereal day filtering (SF), multipath hemispherical map (MHM), and trend surface analysis-based MHM (TMHM) methods. The effects of the MHM method decreased rapidly with decreasing resolution due to the reduced ability to capture high-frequency multipaths. The effects of the TMHM method did not change significantly because trend surface analysis mitigated both low-frequency and high-frequency multipaths. The SF model performed worse than the MHM_0.5 and TMHM_1.0 models due to sparse data sampling. Considering the computational efficiency and model accuracy, we selected the TMHM_1.0 model to mitigate multipath effects in the single-epoch PPP test. The test results indicated that multipath corrections substantially improved the reliability and accuracy of single-epoch PPP-CAR. This is because multipath corrections improved the observation accuracy, which increased the instant accuracy of float ambiguity. The instantaneous wide-lane AR was further improved, subsequently further enhancing the instantaneous narrow-lane AR. The most reliable solution was obtained with multipath corrections, as the same fixing rate of 99.8% was achieved for the uncorrected and corrected solutions. The RMSs of the instant positioning errors in the east, north, and up components decreased from 5.1, 5.3, and 18.5&#xa0;cm to 3.3, 3.6, and 11.2&#xa0;cm, respectively, with improvements of 35.3, 32.1, and 39.5%, respectively, for the uncorrected and corrected solutions.</p>