Hi, I would like to understand how Gemini 3 Pro Preview charges for long context under a few scenarios. I understand that the long context rates kick in at 200k context tokens.
My questions are the following:
-
If a request has 250k context tokens, consisting of 200k input tokens and 50k cached input tokens, will both the input tokens and cached input tokens be charged at the long context rate?
-
If a request has 200k context tokens, consisting of 150k input tokens and 50k cached input tokens, will the request be charged at the long context rate?