GPT-6 Astra in Auto Label: Automate data annotation with plain language prompts

You can automate image annotation with GPT-6 Astra in Auto Label when working on object detection projects, either in the Roboflow app or with Roboflow MCP. The model reads open vocabulary class names, visible text, and the relationship between objects, so a class description like "medicine bottle, but only if the bottle is the one at the very front of a row" can find specific objects in a frame based on what you need. When working in Auto Label, you can provide a class name and an optional description, and GPT-6 Astra draws bounding boxes for every match across the batch.

Learn more about Auto Label