Pro models stuck at 1M context size. Can't switch back to 272k

Support Bug Reports
, ,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

I have used the 1M context option on GPT-5.6 Luna only once and now all my GPT models are stuck at 1M even though I’ve switched back to 272k context window. I tried switching back and forth to ensure this isn’t a UI bug but the context chart still shows a 1M size even though the actual setting shows 272k is selected. I don’t know how much of my pro model budget this has burnt.

Edit: Also checked with Gemini 3.8 Flash. That too shows the context is 1M. Is this just a visual glitch the context stats of the chat has or is it actually using a 1M context in the backend, charging me a 2x on input tokens?

Steps to Reproduce

  1. Use Luna with 1M context once.
  2. Switch back to 272k in another chat.
  3. All GPT models now use 1M forever.

Expected Behavior

Models should use the selected context sizse.

Operating System

Windows 10/11

Version Information

Version: 3.20.21 (user setup)
VS Code Extension API: 1.128.0
Commit: f09fca384ceca23f7bf21f9c23655b162641d740
Date: 2026-09-13T19:02:51.781Z
Layout: IDE
Build Type: Stable
Release Track: Default
Electron: 42.10.0
Chromium: 148.0.7778.280
Node.js: 24.18.1
V8: 14.8.178.38-electron.0
xterm.js: 6.1.0-beta.291
OS: Windows_NT x64 10.0.26200

For AI issues: which model did you use?

GPT-5.6 Luna, Terra, Sol

Does this stop you from using Cursor

No - Cursor works, but with this issue

8

Hey @ahkhanjani, thanks for the report.

The context chart is showing the window the request actually used. After the 1M option is selected once, switching the picker back to 272k does not clear it, so GPT-5.6 keeps running with the 1M window even though the picker shows 272k. We’ve let the team know. There is no in-product control that resets it, so you do not need to keep toggling.

The per-token price is the same for 272k and 1M. The higher rate only applies to prompts that go past 272k tokens. Your recent conversations did not get near that, so this has not added an extra charge.

I’ll post here when there’s an update.

10

Hey @ahkhanjani,

A fix for GPT-5.6 staying on the 1M context window after switching back to 272k has landed and will be included in an upcoming version of Cursor (3.22)

One extra step once you’ve picked up that update: because the earlier 1M choice is still remembered, please select 1M once and then switch back to 272k. That clears it, and the context chart should show 272k from then on across your models, including Gemini.

Let me know if you still see 1M after that.