Download Model Weights

You can download model weights for use in edge deployment.

When you train a model on, or upload model weights to, Roboflow, your model is available for download on your own hardware through Roboflow Inference or a model weights file.

Roboflow does not provide support for downloaded model weights used outside of its ecosystem.

Roboflow provides all users with a commercial license to use all models for on-platform training and on Roboflow's Managed Compute.

Paid plans also receive a limited commercial license for use with self-hosted Inference. An extended commercial license for downloaded model weights used outside of Roboflow's offerings is available on Growth and Enterprise plans.

Learn more about model licensing

Downloading Models with Inference

Roboflow Inference is an open source, scalable system that you can use to directly integrate your model into your application logic, or to run a microservice on your hardware through which you can run your model. Inference is designed for scale: Roboflow uses Inference to power our hosted API which has run hundreds of millions of inferences.

Inference supports running models on CPU and GPU devices, from laptop computers to cloud servers to NVIDIA Jetsons to Raspberry Pis.

When you deploy your model with Inference, your model weights are downloaded onto your hardware for use. Your weights are downloaded when you first run a model and stored locally. Predictions are made using your device's local compute and images are not sent into Roboflow's cloud by default.

To learn more about deploying models with Inference, refer to the Inference documentation.

Downloading Weights for Embedded and Mobile Devices

Downloading model weights is only available on certain paid plans.

Some paid plans also include the ability do download model weights for use on devices that Roboflow does not yet natively support (like Android and the Raspberry Pi AI Kit).

Download Weights Button

From within the Roboflow Platform, simply use the "Download Weights" button on the Versions, Models, or Deployments pages of your Project once you've trained a model and you will receive a PyTorch .ptfile that you can convert for use with embedded devices.

Weights file download is available for Object Detection, Instance Segmentation, Classification and Keypoints Detection projects only.

You can also use the Roboflow Python package to download model weights. To do so, use the following code and substitute your Roboflow API key, project ID, and the version associated witht he model you want to download:

import roboflow

rf = roboflow.Roboflow(api_key="YOUR KEY HERE")
model = rf.workspace().project("PROJECT_ID").version("1").model
prediction = model.download()

Your model weights will be downloaded and available in your local directory as a weights.pt file.