Run vision projects from ChatGPT
͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏
|
|
Find the defects you never labeled
|
|
|
Launch
Catch defects your model has never seen
Anomaly detection models learn from images of good parts and mark what's different, so you don't need labeled examples of every defect. Now you can train, evaluate, and deploy anomaly detection projects in Roboflow.
|
|
| |
|
|
Product update
Label, train, and deploy without leaving ChatGPT and Codex
Install the Roboflow plugin in ChatGPT and run a vision project from plain prompts. Upload and auto-label images, train a model, and save a Workflow, all in one chat.
|
|
| |
|
|
Blog
Bringing real vision tasks to physical AI research
Roboflow is now part of NVIDIA's Cosmos Coalition, a group of teams building open world models for physical AI. Roboflow Labs brings tasks from 2M+ developers and Fortune 500 customers to the research, evaluation, and data behind those models.
|
|
| |
|
|
| |
|
|
New in Playground
GPT-6.1 Terra posts the highest object detection score Roboflow has measured, ahead of Gemini 4 Argon and GPT-6 Astra. Run it side by side with either on your own images.
Try it in Playground →
|
|
| |
|
|
| |
|
Research of the week
|
|
|
| |
|
© 2026 Roboflow, Inc. All rights reserved.
500 Locust St, Suite 104, Des Moines, IA 50309
Manage Preferences |
Unsubscribe
|
|