Are you seeing Roboflow instant model in the dropdown?

We recommend using the new “Active Learning” option on the project; see the left-hand sidebar to find it. As for quickly training new models, RF-DETR nano is what we all use for quick prototyping and label assist when we’re getting started on a new dataset. It’s our go-to for prototyping, then once we think we’re ready we’ll train a larger RF-DETR to deploy to our serverless API, or run it on our own hardware using Roboflow Inference.