Deployment

Try a trained model in the built-in playground, then export it in the format that fits where inference will run.

Try it first

The Deployment panel includes a playground: enter feature values and run inference against a trained model before you export anything.

Export formats

The Deployment panel exports a trained model as a quantized artifact, a static browser bundle, or a server-side wrapper, depending on where you plan to run inference.

What happens to your data

Export is explicit and local: nothing leaves your machine until you choose to download or publish it.

Go deeper

See Deploying and packaging models for validation, the Hugging Face tab, standalone export and pipeline packages, or follow Deploy and Export a Model.