Multi-GNSS PPP Partial Ambiguity Resolution (PPP-PAR) can improve the fixing success rate and time to first fix (TTFF). Although the elevation angle ranking strategy can suppress worse observation quality of low elevation satellite, the observation quality of high elevation satellite still depends on prior weight. The PPP-PAR fixed success rate and TTFF based on elevation angle sorting will decrease because the prior weight cannot reflect the real observation quality. Therefore, we proposed a multi-GNSS PPP partial ambiguity resolution method based on observations weight ranking. Firstly, the wide-lane and narrow-lane cascading PPP-AR based on dual-frequency ionospheric-free (IF) combination model is adopted. The weight of the wide-lane float ambiguity of the anomaly observation epoch in the multi-epoch smoothing filter is reduced, and the nearest rounding decision function is used to fix the between-satellite single-difference (BSSD) wide-lane ambiguity. Thus, the risk of wide-lane ambiguity fixing error caused by observation gross error is reduced. Finally, the partial subset will be determined by posterior IF observation weight ranking, and the narrow-lane integer ambiguity can be fixed based on LAMBDA. The data of 32 MGEX stations in 7 days is collected to verify the proposed method. Compared with the elevation angle priority fixed method using IGG-3 model, the advantages of the proposed method are as follows: the TTFF and convergence time of fixed solution is increased by 8.7 % and 7.2 % respectively, the fixed success rate is increased to 94.9 %, and the RMS of north, east and up is reduced to 6.6 mm, 7.8 mm and 37.2 mm respectively.