The context window in the docs shows “128k” context window, even in MAX mode. Inside of Cursor’s model selector, it shows 1M.
Hey, thanks for the report. We’ll fix it.
Which one is correct?
Gemini 2.5 Flash has a 1M context window, as shown in the model selector.
Hi, just wanted to update that even after the fix, it is still wrong. Both MAX and non-MAX mode are showing as 1M context, which is clearly incorrect.
why would this be a mistake? in the comment above it is confirmed both should be 1M and google docs also confirm they both have 1M context window
This is a mistake because in the IDE, when not using MAX mode, it says that the context window is 128K. It is unlikely imo that we get the full 1M context otherwise long conversations would be really expensive to absorb costs of.
You can see that in MAX mode, only then do you get the full 1M context.
that is correct, in the max mode you pay for the tokens you use
In the IDE it appears like this:
Gemini 2.5 Flash
Fast, low-cost, simple tasks.
128k Context Window
0× Requests
Version: preview-05-20
The error in the docs specifically for the case of non-MAX mode, which says that it has full 1M context, but in the IDE it says that the context is 128K.
So, there 1m context for max mode only or it not only for max mode? Where is mistake? Docs or IDE?
Most likely docs but waiting for official response.
Hey, yes, that’s a mistake, we’ll fix it. Thank you.
The error is in the documentation or in the IDE?
But you’re not answering whether the error is in the documentation or in the IDE.
Hey, sorry, I forgot to mention — this is an error in the documentation: this model has 1M in Max mode and 128K in non-Max mode.