[whatever1]

Frontier labs provide “frozen” builds of their models that hyperscalers just serve without collecting data. This is a prerequisite from most of the companies that store sensitive data and still want to use frontier LLMS.