YOK: GPT-5.6 reasoning content not shown even when returned by OpenAI

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

When using Bring Your Own Key (custom OpenAI Base URL + API key) with GPT-5.6 models, Cursor does not display the reasoning content (thinking steps) even when the OpenAI endpoint returns it.
Even when reasoning effort is manually forced to maximum on the proxy side, Cursor still hides the reasoning content. Other tools (e.g. Warp) correctly show the reasoning content from the same endpoint. This makes it impossible to see the model’s reasoning when using custom OpenAI endpoints.

Steps to Reproduce

Configure Override Base URL + API Key for OpenAI in Settings.
Select a GPT-5.6 model and set reasoning effort to Max.(even though Cursor has another bug there that does not send the reasoning effort to the Endpoint.)
Use a proxy and manually force reasoning_effort to max (or xhigh) in the request.
Send a prompt that normally triggers visible reasoning.
Observe that:
The OpenAI backend returns reasoning content.
Cursor does not show any reasoning content in the UI.
The same proxy + model in Warp (or another client) does show the reasoning content.

Expected Behavior

Cursor should display the reasoning content returned by OpenAI when using a custom base URL, the same way it does with official OpenAI models and other clients like Warp.

Operating System

MacOS

Version Information

Version: 3.11.13
VS Code Extension API: 1.125.0
Commit: 3f21b08f0b436a07be29fbfe00b304fa15553350
Date: 2026-07-10T01:45:28.254Z
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.256
OS: Darwin arm64 27.0.0

For AI issues: which model did you use?

GPT-5.6 Sol / GPT-5.6 Terra with Max reasoning effort

Additional Information

This is a separate but related issue to the missing reasoning_effort parameter in requests. Even when reasoning is forced on the backend, Cursor does not render the reasoning content for custom OpenAI endpoints. Related thread: BYOK does not send the "reasoning_effort" parameter for GPT-5.6 models

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

5

Hey, thanks for the detailed report and for splitting request-side and render-side into separate threads, that really helps.

Confirmed, this is a bug on our side, not an issue with your setup. When Override OpenAI Base URL is set, GPT-5 family requests go through a path where reasoning content reasoning_content or reasoning isn’t mapped into thinking, so the UI shows nothing even though the endpoint returns it and other clients like Warp read it directly. On the default managed OpenAI path, reasoning renders correctly.

I’ve linked this internally with your sibling issue about reasoning_effort BYOK does not send the "reasoning_effort" parameter for GPT-5.6 models since they share the same root cause. I can’t share a fix timeline yet.

Workaround for now: if you need to see reasoning for GPT-5, use the standard OpenAI provider without a custom base URL, rendering works there.

I’ll post an update in the thread when I have one.