It is worth noting that while the G magnitudes are excellent, there are known small issues for bright blue stars and faint red stars. Details can be found in Riello et al 2020.
The bright blue stars are G<13 and BP-RP<0.1, appear to have an ‘anomaly’ of up to a few mmags.
The Gaia web pages provide Python to correct the systematic in fainter G magnitudes, G>13. This is not a major problem and in most cases will be very small except for redder stars. For example it can reach about the 1% level for about BP-RP>2, though at the extreme red end can reach 2.5%. I’m using a few million Gaia sources, and in my sample the official correction adjusted 1% of the stars by 1% in G. There is mention in the paper of a separate table in the Gaia archive for corrected G-band photometry, though I’ve not seen that yet and I wonder if it will appear in the coming days.
There also appears to be parallax zero point bias by magnitude, colour and position (Lindegren et al 2020). Again a Python script is provided to calculate this bias.
If anyone is interested in the Python corrections, then the source code can be found on the Gaia webpages here: https://www.cosmos.esa.int/web/gaia/edr3-code
This is not intended to take away anything from the Gaia results, which are amazing. It is usual for a major survey to have some small issues to be aware of, especially noting this is just an interim data release.