OpenAI
Label Studio Enterprise integrates with OpenAI through the Prompts interface so you can call models like GPT-4.1, GPT-4o, and gpt-4o-mini directly from your labeling projects. Once you add your OpenAI API key, these models appear as selectable providers in Prompts, letting you generate annotations, run structured evaluations, and compare model behavior against ground truth.
With this connection, you can:
- Pre-label tasks using OpenAI models and write results into project controls (Choices, Labels, TextArea, etc.)
- Compare model outputs to existing ground truth and track quality metrics before scaling up
- Use real project data to iterate on prompts, evaluate changes, and roll out auto-labeling with more confidence
This OpenAI integration is available in Label Studio Enterprise via Prompts.
How Label Studio Connects with OpenAI in Prompts
Label Studio Enterprise treats each Prompt as a reusable “model” definition that knows which OpenAI model to call, what prompt template to use, and which project fields to read and write. When you run a Prompt:
- The Prompts UI sends your selected tasks, prompt template, and model choice to the Enterprise backend
- The backend calls OpenAI with your API key, the chosen model, and task data (via our internal orchestration layer)
- Responses are written back into your project as predictions, which you can inspect, edit, and compare to ground truth
- When you’re satisfied with quality, you can use the same Prompt to generate predictions across more tasks for assisted or automated labeling
