Skip to content

9.2. OpenAI Integration

Connect OpenAI's GPT models and use them alongside your local models.

Supported Models

  • GPT-5.4: OpenAI's flagship model, strongest at complex reasoning.

  • GPT-5.2: Optimized for speed and cost-efficiency while maintaining high intelligence.

  • GPT-4o (Legacy): A reliable and cost-effective option for less demanding tasks. (Older models like GPT-4 Turbo are deprecated).

Setup with an API Key

  1. Get your API Key from the OpenAI Platform.

  2. In Backend.AI GO, go to Providers > Add Provider > OpenAI.

  3. Enter your API Key.

Sign in with ChatGPT

If you have a ChatGPT subscription, you can connect OpenAI without an API key. In the provider setup wizard, choose the ChatGPT sign-in option and Backend.AI GO starts a device-code flow:

  1. A dialog shows a one-time code and opens the OpenAI verification page in your browser.
  2. Enter the code on the verification page and authorize Backend.AI GO. The dialog shows how long the code remains valid; if it expires before approval, start the flow again.
  3. Once approved, the provider card shows that you are signed in via Codex OAuth.

The authorization is stored locally per provider, and no API key is created. To disconnect, click Sign out on the provider card.

Usage Tips

  • JSON Mode: OpenAI models support strict JSON output, which works well for tool calling and data extraction tasks in Cowork.

  • Image Input: GPT-5.4 and GPT-5.2 can analyze images. Drag and drop an image into the chat to use this feature.