When instructing Cursor (using Sonnet 4.5) to begin editing a file, an error consistently presents saying: “Unable to reach model provider”. Even when I change the prompt back to something that does not require editing a file, it still will produce the same error from then on and that Agent will be unusable.
Steps to Reproduce
Set model to Sonnet 4.5 (although I did have the same issue with GPT 5.1 as well)
Leave mode set to Agent
Plan for around 100k tokens
Then once happy with the plan, instruct Agent to edit the respective file
Error presents and the chat becomes unusable.
This has been consistent for me, there will be no crash UNTIL I instruct it to edit the file. I always keep the session in Agent in Agent mode.
Expected Behavior
The agent should edit the file without responding with ‘unable to reach model provider’
Operating System
MacOS
Current Cursor Version (Menu → About Cursor → Copy)
Hey, thanks for the report. The “Unable to reach model provider” error has been affecting several users over the past few days when using Sonnet 4.5 and is currently under investigation.
Your specific trigger (editing a file after a large planning step) is very helpful. Could you please share:
Network diagnostics: Cursor Settings → Network → Run Diagnostics (share any errors if they appear)
Request ID: from the chat context menu (top right) → Copy Request ID when the error appears
Network setup: are you using a VPN, corporate network, firewall, or antivirus?
HTTP/2 test:
CMD + , → search for “HTTP/2”
Enable “Disable HTTP/2”
Restart Cursor
Test again in a new Agent chat
New chat behavior: does creating a new Agent chat help temporarily, or does the error appear right away?
Let me know the results and we can dig deeper if the standard steps don’t help.
Request ID: fed9933a-e006-4459-9fc8-2e6be2d97560
{“error”:“ERROR_OPENAI”,“details”:{“title”:“Unable to reach the model provider”,“detail”:“We’re having trouble connecting to the model provider. This might be temporary - please try again in a moment.”,“additionalInfo”:{},“buttons”:,“planChoices”:},“isExpected”:false}
ConnectError: [unavailable] Error
at RZc.$endAiConnectTransportReportError (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:6330:408452)
at GPo._doInvokeHandler (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:7180:11314)
at GPo._invokeHandler (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:7180:11056)
at GPo._receiveRequest (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:7180:9818)
at GPo._receiveOneMessage (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:7180:8635)
at bPt.value (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:7180:6727)
at _e._deliver (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:49:2962)
at _e.fire (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:49:3283)
at Qmt.fire (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:6315:12156)
at MessagePort. (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:8972:18439)
Network setup: no VPN, home network so no firewall/antivirus. running MacOS version 15.6.1 (24G90)
HTTP/2 test: New chat works fine
New Chat behaviour: always seems to be fine at first. The crash this time round for Sonnet 4.5 happened at 249k tokens, but has happened with GPT 5.1 at 150k tokens. I’ve never had it at the beginning of a chat.
Crash continues to happen consistently. I can moderately plan with Cursor but cannot write to any files without the crash occurring - and even planning crashes after a certain (but inconsistent) context window size. Right now I’m essentially just planning with Cursor where possible and then getting Claude Code to execute. Let me know what other details might be helpful, happy to help get to the bottom of this.
i had the same problem… this is nto resolved for me… i raised a topic about thsi as well… request details - Request ID: c9e38da4-03d7-4b53-91b3-f4821198931d
{“error”:“ERROR_OPENAI”,“details”:{“title”:“Unable to reach the model provider”,“detail”:“We’re having trouble connecting to the model provider. This might be temporary - please try again in a moment.”,“additionalInfo”:{},“buttons”:,“planChoices”:},“isExpected”:false}
ConnectError: [unavailable] Error
at m3c.$endAiConnectTransportReportError (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:6330:434848)
at qRo._doInvokeHandler (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7241:11314)
at qRo._invokeHandler (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7241:11056)
at qRo._receiveRequest (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7241:9818)
at qRo._receiveOneMessage (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7241:8635)
at jRt.value (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7241:6727)
at ke._deliver (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49:2962)
at ke.fire (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49:3283)
at Cgt.fire (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:6315:12156)
at MessagePort. (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:9033:18439)
I completely agree with you and was so annoyed by the reduction in productivity, and I expect to be reimbursed in some way for the days that I wasn’t able to use the features that I am paying for. I’ve emailed about that (no response as yet) - but this post specifically wasn’t about that I’m also just acknowledging that the staff are probably working hard to resolve this for us. But yeh we’re certainly on the same team here as I’m sure the staff are too with the frustration.
is there any way we can approach the staff??? mailing just sends it to the effin bot .. sonnet 4 1m works… 4.5 does not.. gemini 3 constantly fails spectacularly
I have been experiencing this for 2 weeks now, seems to cap out at 200k context even with MAX mode enabled… Using Sonnet 4.5, when I start getting the error i change the model to Gemini 3 Pro so it completes the task…
I’m also getting this issue today, after getting it last week. I’ve updated to the latest version of Cursor, too. I’m getting this with both Sonnet 4.5 and Opus 4.5. This seemed to start happening when Opus 4.5 came out, maybe something has broken in the Anthropic model connection? I have to switch away from using Anthropic models until this is fixed, I get this issue almost every time.
