Cursor AI agent constantly reverts to "Auto" model

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

After installing a recent update, the number of times the agent has switched to “Auto” from my selected model is intolerable. New chat sessions and plans sneakily switch the selected model for the session to “Auto”…

It appears this is a known class of issue, but I never had to deal with this until one of the very last few updates, and I’m fairly surprised this is not more prominent of an issue currently

Steps to Reproduce

Often just using a session where you switch the session model to a selected one like “Opus 4.7 thinking 1M high”, and have it create a plan, it will sometimes select “Auto” inside the plan block itself, overriding your prior selected model with which you created the plan itself.

Or when the “Taking a while…” message appears (which has also been appearing more often, though there’s already a report for that), stopping, and resubmitting a prompt.

Expected Behavior

My selected model should remain selected consistently, even in generated plans, or after recovering from interruptions that trigger the “Taking a little longer than expected” message. It’d also be nice if there was a stronger visual indication that the model was changed or that some of the session was done with a different model than the currently selected one.

Operating System

MacOS

Version Information

Version: 3.1.17
VSCode Version: 1.105.1
Commit: fce1e9ab7844f9ea35793da01e634aa7e50bce90
Date: 2026-04-19T19:33:58.189Z
Layout: editor
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Darwin arm64 25.5.0

For AI issues: which model did you use?

Opus 4.7 Thinking 1M High

Does this stop you from using Cursor

No - Cursor works, but with this issue

5

Hey, thanks for the detailed report, especially the two specific scenarios. That’s really helpful.

This is a known class of bugs. In March we shipped a few fixes for resets to Auto, but in 3.1.x there are new paths that can overwrite the selected model, especially in plan mode and with the new backend flow for the initial model. I logged your report internally, but I can’t share a fix timeline yet.

A couple things that would help speed up diagnosis:

  1. The Request ID from one of the sessions where the model switched to Auto inside a plan block (chat > menu in the top-right corner > Copy Request ID).
  2. Can you confirm: is Opus 4.7 Thinking 1M High reliably available in your picker? I’m trying to see if it ever disappears from the list, since that’s a separate regression path.

I’ll also pass along the feedback about a visual indicator when the model changes within a session. That’s a solid idea.