I know at least a couple LLM providers will do some caching for you automatically now, which muddies the waters a bit. [0]

[0] https://developers.googleblog.com/en/gemini-2-5-models-now-s...