Comparing two proportions in dependent samples

#1
Hello,

I am interested in comparing the precision of two classification models, A and B, when I have only the following information:

The number of true positives of model A is TP(A) = 100, the number of false positives is FP(A) = 300.
The number of true positives of model B is TP(B) = 200, the number of false positives is FP(B) = 350.

And precision = TP/(TP+FP).

I do not have any further information (for instance, I don't know which instances were correctly classified by both A and B.)

How do I meaningfully compare the precisions? Is the presicion of B significantly better than that of A?

Since both models used the same data set, the samples are not independent, hence the standard two-proportions z-test is not applicable. Any ideas? Thanks a lot in advance!
Gaston