It sort of depends on which text you're reading, but for the most part they're fine. Some writers knew more about military affairs than others of course (Livy and Plutarch both clearly have absolutely no idea what a battle looks like, although in his books on the Punic Wars Livy's descriptions rely on people who did know, so they're quite good), but in general ancient descriptions are fine. Indeed, with the exception of a few authors like Caesar, who describes military actions in great detail, descriptions of ancient battles tend to be rather sparse. Except when there's really no possible way that it could've turned out that way or we have contradictory evidence there's no good reason to doubt the descriptions of ancient battles that are in our texts. Numbers tend to be inflated, for various reasons, but that's expected and we know to look out for it. I can only think of a handful of battles whose descriptions in our texts don't seem to be accurate--Peter Green makes a somewhat convincing (though I'm not totally sold) argument that the standard description of the Granicus is the result of Alexander tampering with the official record, which he could do because the first few histories of him were written largely by his court historians. Green's argument is that the accounts of the battle from several of our sources don't match and the casualty figures in particular are completely contradictory, and he argues that the initial attempted river-crossing was a costly failure, which would've been very bad for Alexander's reputation, since Philip's advisors (especially Parmenio) had told him not to do it that way. I haven't seen that much discussion about the Granicus, but the other three battles that you list are fine in their historical record. Cannae and Zama are both described by Polybius and Livy, and their accounts match up without significant differences. Gaugamela is described by all our major sources for Alexander's life (Plutarch, Arrian, Diodorus, and Rufus are probably the most important, though of varying quality), and they all read more or less the same. In general when there's no good reason to doubt what's written in the texts we don't