Model switches mid-conversation

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

The model in a conversation switches randomly to a different model mid conversation.

Steps to Reproduce

Repro steps:
When you start a new agent and select a model and have the agent do work, the dialog box in that conversation switches from what it was to something else randomly - leading the user to believe that the original model chosen for the thread is no longer working on the thread AND if you continue the conversation, the model then switches to a new model that you didn’t want. IT IS MADDENING. Please fix it.

Expected Behavior

Expected behavior:
When a model is selected for an agent conversation, do not change the model for that conversation. The user can change it if they want to. The IDE should not do it ever.

Part of this could be a UI issue. We don’t currently have any way to know what model is working at any given time. If you could have some sort of symbol or tag or hover on a given portion of a conversation it would alleviate any fears that cursor is randomly switching the model.

Operating System

MacOS

Version Information

Version: 3.17.21
VS Code Extension API: 1.128.0
Commit: 8f2a112cb2845a97b75fd932ea5c470579ca4060
Date: 2026-08-25T01:05:08.089Z
Layout: Agent Window
Build Type: Stable
Release Track: Default
Electron: 40.10.3
Chromium: 144.0.7559.236
Node.js: 24.15.0
V8: 14.4.258.32-electron.0
xterm.js: 6.1.0-beta.291
OS: Darwin arm64 25.6.0

For AI issues: which model did you use?

I try to use GPT SOL and then it will switch to Opus 5 or the default (And no, this doesn’t only happen when you run out of usage.). Or I’m using Fable and then it switches to Opus 5.

For AI issues: add Request ID with privacy disabled

47789d05-8c49-461e-bd11-3d317ecc32d3

Additional Information

This is really really frustrating. When I choose a model, I expect it to stick. I’m going to have to explore alternatives if you don’t get this under control.

Does this stop you from using Cursor

No - Cursor works, but with this issue

7

Hey @Mike_Johnson, thanks for flagging this!

Quickest fix: Settings → Agents → Conversation → Default Model → “Last Used” (update Cursor if you don’t see it). That keeps whatever model you last picked instead of Cursor choosing for you, and re-selecting the model in the composer right before sending pins it for that chat.

One heads-up: when you hit your usage limit for non-included models, Cursor can auto-switch the chat to a suggested model - a notice shows but it’s easy to miss.

If it still switches on its own after you set Last Used, reply with a fresh request ID and I’ll dig in.

8

Hi Mohit - I already have that setting correct and I confirmed I wasn’t hitting plan limits.

And I just figured out how to reproduce it every time. When I create a plan using model A, and then I hit “Build Locally” on the plan, it switches the model from model A to model B and then starts executing the plan.

Mohit - this is maddening. I’m either wasting a ton of tokens (and money) using a model that is too smart or too dumb for the plan (and I need to redo the plan).

I’m not sure how important you think this is, but for me, this is causing me to doubt investing in the cursor platform. AND I LOVE CURSOR! Please help! :folded_hands:

RequestIds:
e09b7d96-016f-493b-a730-82acd7bd0299
c1bf41b2-93d8-4f4c-811b-d379190af7f8

c26903f0-3036-4eab-bb4d-7ed2c22bdba3