According to the different baseline lengths, we quantitatively analyzed how much precision of the baseline de-termination is improved for GPS survey when using the precise ephemeris instead of the broadcast ephemeris of GPS satellites. For this research, we selected seven baselines ranging from 15 km to 201 km and performed GPS measurements more than six times for each baseline. The observed data for each baseline were processed two times with the same conditions alternately changing the broadcast and the precise ephemeris. The standard deviations from the repeated measurements for each baseline are compared between the results of using the broadcast ephemeris and the precise ephemeris. As the result, the precision, stability and reliability of the base-line determination using the precise ephemeris is better than those of using the broadcast ephemeris for all base-lines. When using precise ephemeris for the baselines longer than 65 km, the precision less than 0.1ppm is always obtained and the precision improvement rate by using the precise ephemeris is considerably greater than that for the shorter baselines. We expect that this result might be a quantative basis for the decision about what ephemeris is better for the baseline length and the demanded precision in GPS survey.