How to use Unquant in Cursor

Before you start

• A paid Cursor plan (Pro or higher)the free Hobby plan saves your key, base URL and custom model but won't let you use them. If Cursor says you need a plan or should upgrade, that message is Cursor's: the request never reaches Unquant.
• The chat panel onlyyour model answers in the chat panel's Ask mode. Ctrl+K inline edits and Tab autocomplete always use Cursor's own models, and custom models in Agent mode have failed on Pro since a Cursor bug in late August 2026.
• The public addressCursor sends these requests from its own servers, so use https://api.unquant.io/v1, not a proxy on your computer.

1. 1

   Turn on your Unquant key

   Open Cursor Settings → Models. Under API Keys, paste your Unquant key into OpenAI API Key and switch its toggle on. Pasting the key is not enough: until the toggle is on, Cursor keeps using its own service.

   Cursor Settings → Models

   OpenAI API KeyYour Unquant API key

   Override OpenAI Base URLhttps://api.unquant.io/v1

   Custom modelglm-5.3-flash
2. 2

   Override the base URL

   Switch on Override OpenAI Base URL and enter https://api.unquant.io/v1.

3. 3

   Add the model

   Click + Add Custom Model, enter glm-5.3-flash (or qwen-3.8-flash-next on a Qwen plan) and make sure it is enabled.

4. 4

   Chat with it

   In the chat panel, choose Ask mode and pick your model from the model menu.

Want a full coding agent?

Cursor keeps its agent, inline edits and autocomplete on its own models. To run a whole agent on Unquant, use OpenCode, Claude Code, or Cline, Roo Code and Kilo Code in VS Code.