Better prompt caching for GPT-6

GPT-6 introduces improved prompt caching with higher cache hit rates, new diagnostics, explicit breakpoints, and controls that reduce latency and costs. Developers can now set breakpoints to manage cache invalidation more precisely, while enhanced diagnostics offer visibility into cache performance. These features aim to optimize efficiency for AI applications, lowering operational expenses and speeding up response times.