70k input 70k output 60k reasoning = 200k, a simple refactor job or documentation request will get to 200k
Context length issues:
Source: [2501.05414] LongProc: Benchmarking Long-Context Language Models on Long Procedural Generation
(Cursor does this periodic summarization when going forward in a chat, which is the way researchers tell us it gives the most quality and as a thank you users in this forum prefer to spam nonsense instead of getting informed)