If you go and buy a class E1 standard weight, the best below having equipment calibrated at NIST or NPL, then the uncertainty in the weight, when weighed in vacuum, is about 0.5 PPM , which corresponds to 21 bit accuracy. +/- 1 bit. This is the best there is, it is not routinely achieved.

To calibrate the equipment to that precision, compensation would have to be made for air density and humidity.

Looking at the Sartorius catalogue, Their jewellers grade balances have an accuracy of 1 part in 250,000

Their chemical balances offer a maximum range of 2 grammes, but a breathtaking reproducibility of 0.25 ugrammes - which, at 0.125 ppm, you will note is better than the uncertainty in the reference standard, so its NOT accuracy. The accuracy is determined by the ability to standardise the instrument.

24 bit which is what Jerry is fighting for is 8 times better !