Why always reset my model list

[AndersRHZ]

Describe the Bug

2.0.63
505046dcfad2acda3d066e32b7cd8b6e2dc1fdc0
arm64

every time i checked the gpt-5 and etc by my perference. cursor will reset my model list after a while.

Steps to Reproduce

1. select my perference models
2. wait for a while
3. then can see my model list have been reset

Expected Behavior

do not reset my model list

Screenshots / Screen Recordings

Operating System

MacOS

Current Cursor Version (Menu → About Cursor → Copy)

2.0.63
505046dcfad2acda3d066e32b7cd8b6e2dc1fdc0
arm64

Does this stop you from using Cursor

No - Cursor works, but with this issue

[deanrie]

Hey, thanks for the report. To help us investigate, please share:

When does the reset happen?

• In the same session, or only after restarting Cursor?

Scope:

• Does this affect all models or just GPT-5?

Any extra details will help me escalate this to engineering.

[AndersRHZ]

1. in the same session
2. i think is all models… i have been check sonnet 4.5 non thinking, haiku4.5-non-thinking, gpt-5-xxx，and etc
3. and sometimes, it only left kimi, and composer. i need to click the refresh button in models list of settings, to restore the basic selected models.

[AndersRHZ]

1. in the same session
2. i think is all models… i have been check sonnet 4.5 non thinking, haiku4.5-non-thinking, gpt-5-xxx，and etc
3. and sometimes, it only left kimi, and composer. i need to click the refresh button in models list of settings, to restore the basic selected models.
4. update, it seems fetch the model list interval from network? it happens the point 3 above…. my model list show only kimi any composer. and suddenly it update to the “default” model list look after…

[AndersRHZ]

any update? how can i fix it ….it annoy…

[AndersRHZ]

here the same， every day every time…auto reset my model list …how can i fix it ?

[AndersRHZ]

again …. reset my model list ,what happen ??

[AndersRHZ]

The problem still exists.

reload window and restart cursor also not work.
uninstall and install a fresh dmg package from download also not work.

[AndersRHZ]

@deanrie do you have any idea ? how can i fix it?

it is very annoy…almost everytime i start a new chat , i have to reselect my favor model….

[deanrie]

Thanks for the info. This is an unusual issue - the model list shouldn’t reset automatically like this.

Before escalating to engineering, could you please share:

1. Current version: Help > About Cursor > Copy (to confirm you’re on latest after reinstall)
2. Console errors: When the reset happens, check Developer Tools > Console tab for any errors
3. API keys: Cursor Settings > Models - do you have any custom API keys configured?

Based on your observation that it seems to “fetch the model list from network,” this could be a configuration conflict.

Let me know what you find, and I’ll escalate this to the engineering team with the additional details.

[AndersRHZ]

Version: 2.0.77
Commit: ba90f2f88e4911312761abab9492c42442117cf0
Date: 2025-11-13T23:10:43.113Z
Browser: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Cursor/2.0.77 Chrome/138.0.7204.251 Electron/37.7.0 Safari/537.36

——

here the error

workbench.desktop.main.js:5827 Element already has context attribute: editor-instance

workbench.desktop.main.js:55  WARN ToolCallEventService: Tracked tool call end - call_bVM3TUUcL0R7Zg7178QEGX2B
fc_0c173162758c212201691c46ed8e8c8198aab13e738bb7eebf

workbench.desktop.main.js:5606 [transport] Stream error reported from extension host ConnectError: [canceled] This operation was aborted
    at QJc.$endAiConnectTransportReportError (workbench.desktop.main.js:4989:399336)
    at Ywo._doInvokeHandler (workbench.desktop.main.js:492:36007)
    at Ywo._invokeHandler (workbench.desktop.main.js:492:35749)
    at Ywo._receiveRequest (workbench.desktop.main.js:492:34514)
    at Ywo._receiveOneMessage (workbench.desktop.main.js:492:33336)
    at GDt.value (workbench.desktop.main.js:492:31429)
    at _e._deliver (workbench.desktop.main.js:49:2962)
    at _e.fire (workbench.desktop.main.js:49:3283)
    at ggt.fire (workbench.desktop.main.js:4974:12156)
    at MessagePort.<anonymous> (workbench.desktop.main.js:7498:18439)

{arch: 'arm64', platform: 'darwin', channel: 'stable', client_version: '2.0.77', streamId: 'd5f60c77-503c-4658-a3b2-a5f420fbc0d1', …}

workbench.desktop.main.js:2750 [composer] Failed to get complete AI response

workbench.desktop.main.js:5606 [composer] Chat incomplete or aborted

{arch: 'arm64', platform: 'darwin', channel: 'stable', client_version: '2.0.77', requestId: '6558e054-b7f0-4828-85c7-eedd59f0f02a', …}

workbench.desktop.main.js:5797 [AiCodeTracking] Error scoring commit 97b399de9b6cf5a05e88e661519cacf43e7d2a41: RangeError: stdout maxBuffer length exceeded

workbench.desktop.main.js:5797 [AiCodeTracking] Error scoring commit 6f3a7d98531257830b91e0e0cfd7748f00377b6a: RangeError: stdout maxBuffer length exceeded

workbench.desktop.main.js:55  WARN ToolCallEventService: Tracked tool call start - call_M9j2g46UC3nbXLkoFUtwUjSZ
fc_0d36747bd6548cac01691c485418b8819797b4ca53f2a2a9c3 (create_plan)

workbench.desktop.main.js:55  WARN [ExtensionService] No inspect port found for LocalWebWorker

workbench.desktop.main.js:5606 [composer] Extension host became UNRESPONSIVE undefined

{arch: 'arm64', platform: 'darwin', channel: 'stable', client_version: '2.0.77', extensionHostKind: 'LocalWebWorker', …}

workbench.desktop.main.js:55  WARN [ExtensionService] No inspect port found for LocalWebWorker

workbench.desktop.main.js:5606 [composer] Extension host became UNRESPONSIVE undefined

{arch: 'arm64', platform: 'darwin', channel: 'stable', client_version: '2.0.77', extensionHostKind: 'LocalWebWorker', …}

workbench.desktop.main.js:55  WARN [ExtensionService] No inspect port found for LocalWebWorker

workbench.desktop.main.js:5606 [composer] Extension host became UNRESPONSIVE undefined

{arch: 'arm64', platform: 'darwin', channel: 'stable', client_version: '2.0.77', extensionHostKind: 'LocalWebWorker', …}

workbench.desktop.main.js:55  WARN ToolCallEventService: Tracked tool call end - call_M9j2g46UC3nbXLkoFUtwUjSZ
fc_0d36747bd6548cac01691c485418b8819797b4ca53f2a2a9c3

workbench.desktop.main.js:5606 [transport] Stream error reported from extension host ConnectError: [canceled] This operation was aborted
    at QJc.$endAiConnectTransportReportError (workbench.desktop.main.js:4989:399336)
    at Ywo._doInvokeHandler (workbench.desktop.main.js:492:36007)
    at Ywo._invokeHandler (workbench.desktop.main.js:492:35749)
    at Ywo._receiveRequest (workbench.desktop.main.js:492:34514)
    at Ywo._receiveOneMessage (workbench.desktop.main.js:492:33336)
    at GDt.value (workbench.desktop.main.js:492:31429)
    at _e._deliver (workbench.desktop.main.js:49:2962)
    at _e.fire (workbench.desktop.main.js:49:3283)
    at ggt.fire (workbench.desktop.main.js:4974:12156)
    at MessagePort.<anonymous> (workbench.desktop.main.js:7498:18439)

{arch: 'arm64', platform: 'darwin', channel: 'stable', client_version: '2.0.77', streamId: '20e913e7-299a-4565-8434-89c68edc0cc6', …}

workbench.desktop.main.js:2750 [composer] Failed to get complete AI response

workbench.desktop.main.js:5606 [composer] Chat incomplete or aborted

{arch: 'arm64', platform: 'darwin', channel: 'stable', client_version: '2.0.77', requestId: '8d10233b-4cbc-49e3-abcf-a16f56299582', …}

workbench.desktop.main.js:5794 Failed to flush analytics events: ConnectError: [unavailable] HTTP 504
    at t (workbench.desktop.main.js:4989:400501)
    at async Object.batch (workbench.desktop.main.js:590:71574)
    at async KKr.flushAll (workbench.desktop.main.js:5794:1142)

—–

no custom api keys

[deanrie]

Thanks for the detailed console logs. The errors show network connectivity issues (HTTP 504 errors, transport failures, and the extension host becoming unresponsive) that seem to be triggering unwanted model list refreshes from the server.

Before I escalate this to engineering, please try the following:

• Disable HTTP/2: open Cursor Settings CMD + , > search for “HTTP/2” > enable “Disable HTTP/2” > restart Cursor

This often fixes intermittent network issues that can cause config sync conflicts.

Let me know if disabling HTTP/2 stops the model list from resetting. If not, I’ll pass this on to the engineering team with your diagnostic info.

[AndersRHZ]

disable http2 not work…