On May 3, 2012, the Jinta M_S5.4 earthquake occurred beneath the northeast margin of the Tibetan Plateau, Gansu, China. We analyzed the broadband waveforms for the mainshock with sufficient signal-to-noise levels to invert for seismic moment tensors. All stations involved in this study are broadband stations situated near the May 3 mainshock; data showing such problems as misorientation of horizontal components, mislabeling and polarity reversal in one or more components were removed. During the data processing, we removed the mean value and instrument response, detrended the waveforms, integrated the three-component velocity waveforms to displacement, and rotated the horizontal components to the radial and tangential directions. Before performing moment tensor inversion, waveforms were filtered with a Butterworth Filter with frequencies between 0.016 Hz and 0.05 Hz. On the basis of the Crust 1.0 global crustal model, and considering the heterogeneity of the crust and the distances to the recording stations, various velocity models for each station were employed to account for variabilities in the crustal structure. During the inversion process, to insure the quality of the results, we retained only the station with variance reduction(V_R) is ≥ 80, and got 5 stations for the inversion. After depth iteration, the best solutions obtained at a depth of 12 km were 71, 29, and 91.5% for double couple(DC), the compensated linear vector dipole(CLVD) component, and V_R, respectively, which indicates high-quality results. The two panel's strike, dip, and rake of the solution were 166°, 66°,-168° and 71°, 79°,-25° respectively. It is found that our inversion results are basically consistent with the strike of the Shenluoshan-Yemajing fault, and the aftershocks distribution occurred near the fault. Associating with the tectonic framework, damage characteristics, and the aftershock distribution feature, we suggest that Jinta M_S5.4 earthquake was caused by the combined action of southeast Mashan fault and Shenluoshan-Yemajing fault, and the latter should be the seismogenic fault of the earthquake.