Summarization does not reduce context, 272k limit is ignored, and context indicator stays at 1M

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Summarization does not reduce context size, even if I run it twice in a row. The selected context limit is also not respected. I set GPT 5.5 to 272k, but the session still went up to around 350k. The context indicator also stays at 1M max instead of reflecting the limit I selected.

Steps to Reproduce

  1. Open Cursor
  2. Start a chat with GPT 5.5
  3. Set the context limit to 272k
  4. Keep using the chat until the context gets large
  5. Run summarization
  6. Run summarization again
  7. The context size does not go down in any meaningful way
  8. Keep chatting and the session goes past the 272k limit. In my case it reached around 350k.
  9. Also the context meter still shows 1M max instead of matching the limit I set

Expected Behavior

If I run summarization, the context should get smaller.

If I set the limit to 272k, the session should stay within that limit.

The context indicator should reflect the limit I selected and the real usage, instead of staying at 1M max.

Operating System

Linux

Version Information

Version: 3.2.16
VSCode Version: 1.105.1
Commit: 3e548838cf824b70851dd3ef27d0c6aae371b3f0
Date: 2026-04-28T21:07:47.682Z
Layout: editor
Build Type: Stable
Release Track: Default
Electron: 41.3.0
Chromium: 146.0.7680.188
Node.js: 24.15.0
V8: 14.6.202.33-electron.0
OS: Linux x64 7.0.2-arch1-1

For AI issues: which model did you use?

GPT 5.5

Additional Information

This already wasted a few percent of my usage.

Does this stop you from using Cursor

No - Cursor works, but with this issue

7

Hi @opdoeq, thanks very much for reporting!

This is exactly the type of issue where disabling privacy mode briefly and recreating it would make a huge difference. If you’re up for it:

  1. Head to Cursor Settings → General → Privacy Mode and toggle to Share Data (you can switch it right back on after)

  2. Reproduce the issue

  3. Click the three-dot menu on the relevant message and copy the Request ID

  4. Drop it in this thread

Thanks again — this’ll help us track it down a lot faster!!