NobleTrainer

NobleTrainer is no longer being maintained. We now recommend Ultralytics Hub for training custom models for use with NobleAIM.

Why this changed

When NobleTrainer launched, Ultralytics Hub did not yet support the full YOLO11 / YOLO26 detect, pose, and segment workflow that NobleAIM relies on. That has changed: Ultralytics Hub is now the most actively maintained training environment for these architectures, and it is built and supported by the same team that maintains the YOLO models themselves.

Continuing to ship a second, parallel training application stopped being a net benefit to users. Pointing customers at the canonical tool is the better experience.

How to train custom models for NobleAIM today

Use Ultralytics Hub to upload your dataset, train a YOLO11 or YOLO26 detect / pose / segment model, and export the trained weights as an ONNX file.

Drop the ONNX file into NobleAIM and use the built-in Engine Builder to convert it into a TensorRT, ONNX-CUDA, DirectML, or OpenVINO runtime engine that NobleAIM can load directly. The whole pipeline takes minutes once you have a dataset.

Existing NobleTrainer installs

Existing NobleTrainer installations continue to function, but the application will not receive further updates. Models trained with NobleTrainer continue to work with NobleAIM. New users should start with Ultralytics Hub.

Models in the Noble Marketplace are unaffected and remain available to every NobleAIM customer in-app.

Train your next model on Ultralytics Hub

Hub has a free tier suitable for most aim-assist model development. NobleAIM’s Engine Builder handles the ONNX-to-runtime step on your machine.