How do you get rid of losing the precision float in opengl?

In an attempt to model 3D a planet with different levels of detail, they receive the following:
The radius of 6,700 units is naturally divided into quadTree, although it is not important. Thus, the surface of the land is approximately 6,700 from the SC centre, and the closer the camera approachs the surface, the higher the peaks are shaken, which is caused by the loss of precision in the construction of the MVP matrix, as the position of the chamber and the objects is based on high values.
To avoid this, he tried before the Model and View matrix generation to make a preliminary offset of the position of the chamber. I mean, the View of the camera will be like 0.00 and the position of the other obstructions (the sack.xyz position is the camera position.xyz), but in this case, each individual has to regenerate the Model Matrix(s) for each bag, which is too costly for their large number. Is there any haki to resolve the problem and not to resort to such a costeffective operation for human resources?

I don't know exactly how you're building a matrix, but it's possible that its normalization after construction (or preconstruction data). If the matrix is very close, the calculation of the scallar works will result in a loss of accuracy if the matrices are not large, the accuracy is lost somewhere in its construction (possibly because of unnormated baseline data). Removing the position you've been writing about is only part of the rule, it's still necessary to change the size of the multiplication to a suitable standard scallar.