Processing the differential inter-system biases (DISBs) correctly is critical for multi-GNSS fusion positioning. A pivot satellite can be selected for RTK positioning if we can process DISBs correctly, which is the so-called tightly combination between systems. Due to the different frequencies of GPS and BDS, DISB cannot be calibrated in advance. In this paper, a tightly combined BDS/GPS method for RTK positioning model is proposed. A zero baseline is selected to verify the near time-constant characteristic of DISB. In addition, aiming at the limitation of traditional MW combination that the fixed success rate of WL ambiguity is greatly affected by pseudorange noise and has long convergence time, we could use the triple-frequency data to fix a long-wavelength EWL ambiguity. Then assists in solving WL ambiguity. The results show that the fractional part of phase DISB with WL combination is rather stable. Under the medium baseline, the fixed success rate of WL ambiguity floating point solution fraction within 0.2 cycles is 100%, and WL ambiguity can be fixed correctly in the first epoch. Compared with the classical differencing model, the inter-system differencing model can effectively improve the positioning accuracy and reliability, especially for severely obstructed environment. In the normal observation environment, the positioning accuracy can reach 2-3cm in horizontal and 5cm in vertical under the medium baseline. When the number of available satellites is seven, the positioning accuracy can be increased by more than 20%.