The Wide Field Survey Telescope (WFST) is a dedicated time-domain multi-band (u, g, r, i, and z) photometric survey facility under construction. In this paper, we present a preliminary study that assesses the quality of photometric redshifts based on WFST by utilizing mock observations derived with the galaxy catalog in the COSMOS/UltraVISTA field. We apply the template fitting technique to estimate photometric redshifts by using the ZEBRA photometric-redshift code and adopting a modified set of adaptive templates. We evaluate the bias (median relative offset between the output photometric redshifts and input redshifts), normalized median absolute deviation (σ
NMAD) and outlier fraction (f
outlier) of photometric redshifts in two typical WFST observational cases, the single 30 s exposure observations (hereafter shallow mode) and co-added 50 minutes exposure observations (hereafter deep mode). We find bias ≲0.006, σ
NMAD ≲ 0.03, and f
outlier ≲ 5% in the shallow mode and bias ≈0.005, σ
NMAD ≈ 0.06, and f
outlier ≈ 17%–27% in the deep mode, respectively, under various lunar phases. Combining the WFST mock observational data with that from the upcoming CSST and Euclid surveys, we demonstrate that the z
phot results can be significantly improved, with f
outlier ≈ 1% and σ
NMAD ≈ 0.02.