Connect DeepSeek V4 Pro and Any Third-Party Model to Claude Desktop

Enable Developer Mode in Claude Desktop, configure a third-party gateway, and connect supported models such as DeepSeek V4 Pro with a Base URL and API key.

Quick start: turn on Developer Mode in Claude Desktop, open Third Party Inference, then enter the Base URL, API key, and model name provided by your gateway. After relaunch, the model is ready to use.
What you need first

Claude Desktop installed, preferably before signing in

A valid Base URL and API key from the same platform account

The exact model name that your token can access

1. Open Claude Desktop

Start the desktop app and keep it in the normal local client flow. The third-party setup is easier to find there than in the browser version.

2. Enable Developer Mode

Open Help → Troubleshooting → Enable Developer Mode, then wait for the app to relaunch.

If you do not see this option, confirm that you are using the desktop application rather than the web app.

3. Open Third Party Inference

After relaunch, the top menu should show Developer. Open Developer → Configure Third Party Inference.

4. Enter the gateway details

Choose Gateway mode, then enter the Base URL and API key from the same provider account. In this example, the gateway is https://superaiapi.com.

5. Enter a supported model name

You can use a model such as gpt-5.4. Other supported names may include MiniMax-M2.7, kimi-k2.6, glm-5.1, or Claude-family models when the provider exposes them.

6. Apply and relaunch

Click Apply Locally → Relaunch Now. Once the app restarts and the model loads correctly, the connection is complete.

Where to get the token

If you still need a token, log in to superaiapi.com, review model groups and pricing, top up the wallet if needed, and create a token from the token management page.

Compare model groups before creating the token. Lower-multiplier groups are cheaper, while higher-multiplier groups are often more stable.

Final check

After the token is created, open “Use Token”, copy the integration details, and verify that Claude Desktop can start a real conversation with the configured model.