They do, but if you look at the graphs...what is the point of the large context window if accuracy drops off waaaaay before context window is maxed?

This says 400K context window: https://platform.openai.com/docs/models/compare?model=gpt-5

They say:

In the API, all GPT‑5 models can accept a maximum of 272,000 input tokens and emit a maximum of 128,000 reasoning & output tokens, for a total context length of 400,000 tokens.

So it's only 270k for input and 400k in total considering reasoning & output tokens.