As far as I can tell Gemini caching DOES work like OpenAI - see implicit caching here: https://ai.google.dev/gemini-api/docs/caching

I confirmed this by running a bunch of prompts through Gemini 3.5 Flash without doing anything special to configure caching and noting that it comes back with a "cachedContentTokenCount" on many of the responses.

The "storage price" quoted is for an optional Gemini feature that most people don't care about: https://ai.google.dev/gemini-api/docs/caching#explicit-cachi...