Summarizing conversations does not inherit model settings

[ngochip]

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

It seems Cursor isn’t inheriting the model’s settings when summarizing.

I’ve run out of quota, so I’ve been using the custom API.

However, the summary always fails because the model isn’t available in the slow pool???
This error has been going on for about a month, but Cursor doesn’t seem to be aware of it.

workbench.desktop.main.js:34329 [composer] Error during NAL summarization: FAi: Slow Pool Error Claude 4.6 Opus is not currently enabled in the slow pool. Please select another model, or [enable usage-based pricing](command:cursorai.action.showUsagePricingModal) to get more fast requests.
    at P5A (workbench.desktop.main.js:34267:23670)
    at B5A (workbench.desktop.main.js:34267:22658)
    at $5A (workbench.desktop.main.js:34268:6285)
    at Wou.run (workbench.desktop.main.js:34268:10400)
    at async $Oa.summarize (workbench.desktop.main.js:46832:19568)
    at async n5u.summarize (workbench.desktop.main.js:46886:18975)
    at async bMe.triggerManualSummarization (workbench.desktop.main.js:34329:26277)
Caused by: ConnectError: [resource_exhausted] Error
    at llb.$endAiConnectTransportReportError (workbench.desktop.main.js:43803:43022)
    at ZDt._doInvokeHandler (workbench.desktop.main.js:46640:23432)
    at ZDt._invokeHandler (workbench.desktop.main.js:46640:23174)
    at ZDt._receiveRequest (workbench.desktop.main.js:46640:21806)
    at ZDt._receiveOneMessage (workbench.desktop.main.js:46640:20623)
    at j2n.value (workbench.desktop.main.js:46640:18650)
    at Qe._deliver (workbench.desktop.main.js:55:2962)
    at Qe.fire (workbench.desktop.main.js:55:3283)
    at VAn.fire (workbench.desktop.main.js:43785:12180)
    at MessagePort.<anonymous> (workbench.desktop.main.js:49776:18755)

Steps to Reproduce

1. Reach the Cursor limit.

2. Customize Models settings (Open AI Custom Endpoint)

3. Start any conversation

4. Select a conversation summary

Expected Behavior

The conversation should be summarized.
The user’s credentials can be used directly to do this.

Operating System

MacOS

Version Information

Version: 2.6.19
VSCode Version: 1.105.1
Commit: 224838f96445be37e3db643a163a817c15b36060
Date: 2026-03-12T04:07:27.435Z
Build Type: Stable
Release Track: Early Access
Electron: 39.4.0
Chromium: 142.0.7444.265
Node.js: 22.22.0
V8: 14.2.231.22-electron.0
OS: Darwin arm64 25.3.0

For AI issues: which model did you use?

Opus 4.6

Does this stop you from using Cursor

No - Cursor works, but with this issue

[deanrie]

Hey, thanks for the detailed report and the stack trace. This is a confirmed bug on our side.

Right now, the summarization feature doesn’t pass through your custom API credentials. So when you’re out of quota, it falls back to the default pool instead of using your BYOK endpoint. That’s why you’re seeing the slow pool error with Opus.

The team knows about this issue. There’s no ETA for a fix yet, but your report helps us prioritize it.

As a workaround for now, you can switch summarization to a model that’s available in the slow pool, like Sonnet. I know that’s not ideal.

Let me know if you have any questions.