Persistent “Unable to reach the model provider” error (Sonet 4 with Reasoning & Max Mode)

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

sonnet 4 reasoning max mode error. for weeks!

Steps to Reproduce

{
“error”: “ERROR_OPENAI”,
“details”: {
“title”: “Unable to reach the model provider”,
“detail”: “The system failed to establish a stable connection with the model provider. This is not a client-side connectivity issue; the request is being initiated correctly, but the provider is not responding as expected.”,
“additionalInfo”: {},
“buttons”: ,
“planChoices”:
},
“isExpected”: false
}
Request ID: 9cfc32c8-b796-474c-99f9-0e2744fbf049

This issue is not related to local connectivity, client request formatting, or network instability. All client-side layers (network, TLS, API key, and request payload) are functioning normally. The request is successfully dispatched; however, the model provider consistently fails to respond.

This strongly indicates a provider-side service accessibility issue, likely related to endpoint availability, routing, or capacity allocation for Sonet 4 (Reasoning/Max mode).

I would appreciate your assistance in investigating the following:

Whether there are known service disruptions or rate-limiting conditions affecting Sonet 4 with Reasoning or Max Mode.

If the request is being systematically rejected upstream (proxy, gateway, or model distribution layer).

Whether this issue is tied to specific account-level or geographic provisioning constraints.

Looking forward to your technical insights and resolution.

Expected Behavior

{
“error”: “ERROR_OPENAI”,
“details”: {
“title”: “Unable to reach the model provider”,
“detail”: “The system failed to establish a stable connection with the model provider. This is not a client-side connectivity issue; the request is being initiated correctly, but the provider is not responding as expected.”,
“additionalInfo”: {},
“buttons”: ,
“planChoices”:
},
“isExpected”: false
}
Request ID: 9cfc32c8-b796-474c-99f9-0e2744fbf049

This issue is not related to local connectivity, client request formatting, or network instability. All client-side layers (network, TLS, API key, and request payload) are functioning normally. The request is successfully dispatched; however, the model provider consistently fails to respond.

This strongly indicates a provider-side service accessibility issue, likely related to endpoint availability, routing, or capacity allocation for Sonet 4 (Reasoning/Max mode).

I would appreciate your assistance in investigating the following:

Whether there are known service disruptions or rate-limiting conditions affecting Sonet 4 with Reasoning or Max Mode.

If the request is being systematically rejected upstream (proxy, gateway, or model distribution layer).

Whether this issue is tied to specific account-level or geographic provisioning constraints.

Looking forward to your technical insights and resolution.

Operating System

Windows 10/11

Current Cursor Version (Menu → About Cursor → Copy)

Cursor 2.1idle windows 11

For AI issues: which model did you use?

Sonnet 4

For AI issues: add Request ID with privacy disabled

9cfc32c8-b796-474c-99f9-0e2744fbf049

Does this stop you from using Cursor

Yes - Cursor is unusable

Hey, thanks for the report.

Please try the following:

  • Settings > Network > Run Diagnostics: send a screenshot.
  • App Settings Ctrl+, > HTTP/2: enable “Disable HTTP/2”.
  • Ensure there is no VPN/proxy/Zscaler.
  • Send the Request ID + time (UTC).

Additional diagnostics would help:

  • Try the same requests on Sonnet 4 without Reasoning/Max, and on other models (Sonnet 4.5 / GPT‑5) - let us know which work and which give the same error.
  • If using your own API key, specify the provider.
  • Try a different network (mobile hotspot) to rule out routing issues, and specify the country/region and ISP.
  • Specify the exact time of the error in UTC and whether there were any successful requests to Sonnet 4 Reasoning/Max in recent days.

Your description suggests a provider-side issue specifically for the Reasoning/Max route, but we need diagnostics to check for possible restrictions/routing/capacity issues.

Hi deanrie,

[2025-11-30T00:17:50.327Z] Host: api2.cursor.sh
[2025-11-30T00:17:50.327Z] Servers: 192.168.1.1
[2025-11-30T00:17:50.327Z] Resolved to 23.21.2.40 in 15ms
[2025-11-30T00:17:50.331Z] Resolved to 23.21.2.40 in 1ms
[2025-11-30T00:17:50.331Z] Resolved to 23.21.2.40 in 0ms
[2025-11-30T00:17:50.332Z] Resolved to 23.21.2.40 in 0ms
[2025-11-30T00:17:50.343Z] Host: api2.cursor.sh
[2025-11-30T00:17:50.343Z] Servers: system
[2025-11-30T00:17:50.343Z] Resolved to 13.217.125.230, 34.194.81.213, 35.171.135.63, 3.228.241.206, 23.22.5.222, 23.21.2.40, 3.214.32.76, 52.0.10.195 in 10ms
[2025-11-30T00:17:50.343Z] Resolved to 13.217.125.230, 34.194.81.213, 35.171.135.63, 3.228.241.206, 23.22.5.222, 23.21.2.40, 3.214.32.76, 52.0.10.195 in 0ms
[2025-11-30T00:17:50.343Z] Resolved to 13.217.125.230, 34.194.81.213, 35.171.135.63, 3.228.241.206, 23.22.5.222, 23.21.2.40, 3.214.32.76, 52.0.10.195 in 1ms
[2025-11-30T00:17:50.344Z] Resolved to 13.217.125.230, 34.194.81.213, 35.171.135.63, 3.228.241.206, 23.22.5.222, 23.21.2.40, 3.214.32.76, 52.0.10.195 in 0ms
[2025-11-30T00:17:50.344Z] Result: true

[2025-11-30T00:17:50.311Z] Start
[2025-11-30T00:17:50.654Z] Host: api2.cursor.sh
[2025-11-30T00:17:50.654Z] Protocol: h2
[2025-11-30T00:17:50.654Z] Result: true in 343ms

[2025-11-30T00:17:50.311Z] Start
[2025-11-30T00:17:50.815Z] URL: https://api2.cursor.sh/
[2025-11-30T00:17:50.815Z] Status: 200
[2025-11-30T00:17:50.815Z] IP: 23.21.2.40
[2025-11-30T00:17:50.815Z] Issuer: C=US; O=Amazon; CN=Amazon RSA 2048 M01
[2025-11-30T00:17:50.815Z] Name: api2.cursor.sh
[2025-11-30T00:17:50.815Z] AltName: DNS:api2.cursor.sh, DNS:prod.authentication.cursor.sh, DNS:*.api2.cursor.sh
[2025-11-30T00:17:50.815Z] DNS Time: 11ms
[2025-11-30T00:17:50.815Z] Connect Time: 164ms
[2025-11-30T00:17:50.815Z] TLS Time: 162ms
[2025-11-30T00:17:50.815Z] Result: true in 504ms

[2025-11-30T00:17:50.311Z] Start
[2025-11-30T00:17:50.476Z] Result: true

[2025-11-30T00:17:50.311Z] Sending ping 1
[2025-11-30T00:17:50.474Z] Response: ‘ping’ in 163ms
[2025-11-30T00:17:50.474Z] Sending ping 2
[2025-11-30T00:17:50.623Z] Response: ‘ping’ in 149ms
[2025-11-30T00:17:50.623Z] Sending ping 3
[2025-11-30T00:17:50.769Z] Response: ‘ping’ in 146ms
[2025-11-30T00:17:50.769Z] Sending ping 4
[2025-11-30T00:17:50.917Z] Response: ‘ping’ in 148ms
[2025-11-30T00:17:50.917Z] Sending ping 5
[2025-11-30T00:17:51.063Z] Response: ‘ping’ in 146ms
[2025-11-30T00:17:51.063Z] Result: true

[2025-11-30T00:17:50.312Z] Starting stream
[2025-11-30T00:17:50.477Z] Response: ‘foo’ in 164ms
[2025-11-30T00:17:51.476Z] Response: ‘foo’ in 999ms
[2025-11-30T00:17:52.476Z] Response: ‘foo’ in 1000ms
[2025-11-30T00:17:53.475Z] Response: ‘foo’ in 999ms
[2025-11-30T00:17:54.475Z] Response: ‘foo’ in 1000ms
[2025-11-30T00:17:55.476Z] Result: true

[2025-11-30T00:17:50.313Z] Starting stream
[2025-11-30T00:17:50.313Z] Pushing first message
[2025-11-30T00:17:50.477Z] Response: ‘foo’ in 164ms
[2025-11-30T00:17:50.978Z] Pushing next message
[2025-11-30T00:17:51.118Z] Response: ‘foo’ in 641ms
[2025-11-30T00:17:51.619Z] Pushing next message
[2025-11-30T00:17:51.793Z] Response: ‘foo’ in 675ms
[2025-11-30T00:17:52.295Z] Pushing next message
[2025-11-30T00:17:52.436Z] Response: ‘foo’ in 643ms
[2025-11-30T00:17:52.951Z] Pushing next message
[2025-11-30T00:17:53.092Z] Response: ‘foo’ in 656ms
[2025-11-30T00:17:53.092Z] Result: true

[2025-11-30T00:17:50.310Z] Host: marketplace.cursorapi.com
[2025-11-30T00:17:50.936Z] Response in 626ms
[2025-11-30T00:17:50.936Z] Response: 200
[2025-11-30T00:17:50.936Z] Response Type: cors
[2025-11-30T00:17:50.936Z] Server: null
[2025-11-30T00:17:50.936Z] Result: OK in 626ms

Time: 2025-11-28 20:43 UTC
Location: Turkey, Gaziantep — ISP: Turkcell Superonline
VPN/Proxy/Zscaler: Not Used
HTTP/2: Disabled HTTP/2 enabled (I tried, same result)

Model test results:

  • Sonnet 4 Normal: Working
  • Sonnet 4 Reasoning/Max: Failing
  • Sonnet 4.5: Working
  • GPT-5: Working

Thanks for the detailed logs - they’re super helpful. This looks like a provider-side issue specifically on the Sonnet 4 route in Reasoning/Max modes (a similar problem recently occurred with Sonnet 4.5 and was fixed on the backend: Reproducible - Unable to reach model provider - Sonnet 4.5).

We need a few more data points to escalate this to the engineers:

  • 2-3 fresh Request IDs for failed Sonnet 4 Reasoning/Max requests + exact UTC time of each
  • Are you using your own API key? If yes - which provider?
  • Tests on modes:
  • Sonnet 4.5 Reasoning (without Max) - works or fails?
  • Sonnet 4.5 Max (without Reasoning) - works or fails?
  • Test with a different connection (e.g., mobile hotspot) - results and ISP
  • Confirm you restarted Cursor and tried a new chat after switching “Disable HTTP/2”

Temporary workaround: use Sonnet 4.5, or GPT-5 - you noted they work. Send the above info, and we’ll escalate routing checks with the provider immediately.

I’ve already provided these. I tried a different IP address. Same problem. I’m not using a VPN. I provided the exact UTC time above.

This topic was automatically closed 22 days after the last reply. New replies are no longer allowed.