Recall that when LLMs were first coming out, there was a very loud and aggressive stance *against* using the standard models.

Their choice of which LLM to use is probably related to those earlier, more cautious stances.