The distribution and comparison of different time scales such as Coordinated Universal Time (UTC) are based on GNSS measurements as a standard technique, today. Typically, a precise point positioning (PPP) approach is used to compare the clock error estimates of the GNSS receivers involved, and thereby assess the relative instability between the time scales. Within the framework of the collaborative research center TerraQ we want to explore the minimum achievable frequency instability with modern GNSS equipment. For this purpose, we conducted a dedicated experiment at Physikalisch-Technische Bundesanstalt (PTB), Germany’s national metrology institute in the spring of 2021. We used four state-of-the-art receivers, namely two JAVAD OMEGA and two Septentrio PolaRx5TR, as well as two Leica AR20 antennas. The receivers recorded multi-GNSS observations and were connected to the locally generated approximation of UTC, referred to as UTC(PTB), at all times during the experiment. The GNSS data are analyzed by means of PPP as well as by forming carrier phase single differences (SDs) between two receivers each to investigate the current technical limits of GNSS frequency transfer (FT). For both techniques, we use our in-house developed MATLAB-based software to evaluate the FT performance for three receiver combinations: two consisting of the same receiver type each (intra-receiver), and one with different types (inter-receiver). We also examine the impact of receiver clock modeling (RCM) on the results. In all cases, frequency instability is assessed by means of the modified Allan deviation. Our results show that intra-receiver pairs achieve better results than inter-receiver links. Which observation type and which GNSS (GPS or Galileo) is used, does not have a significant impact on the resulting frequency instability. Furthermore, RCM in PPP shortens the averaging time to reach a certain instability range, e.g., 10^?17 after roughly 30,000–60,000 s, whereas without RCM we do not get to this level at all based on the used ten-day data set. In conclusion, the use of state-of-the-art GNSS receivers should not be a limiting factor when aiming for 5 · 10^?17 instability in frequency transfer.