Segment Anything 3 is out and you can start using it in Roboflow right now. This is the missing link for vision to be as easy to use as LLMs. Completely changes how people will build computer vision applications. If you're in the visual intelligence arena, you have to get your hands on it. Any vision project that you couldn't get to work in the past is probably unlocked by this model. It's fully integrated into the Roboflow platform as of today. We have a dedicated API endpoint for you to use, we have the fine-tuning infra built so you can train it on your own data, it's in Inference for you to deploy locally or in your own cloud, and you can use it to label data for training smaller models. All of that is available to all users right now. So many new vision applications just became possible. 2026 is going to be the year of visual intelligence. Big thanks to Meta for making this model open and available for use.