Hugging Face adds one-click SageMaker Studio deploy button
Hugging Face model cards now include a direct deploy-to-SageMaker Studio button, letting AWS users spin up inference endpoints without leaving the browser.
Hugging Face rolled out a one-click deploy button that sends models straight to Amazon SageMaker Studio. The button appears on model cards across the Hub and launches a pre-configured SageMaker endpoint in the user's AWS account. The feature targets practitioners who want to move from browsing open weights to production inference without writing deployment scripts.
The integration uses SageMaker's JumpStart catalog under the hood. When a user clicks the button, Hugging Face passes model metadata to AWS, which provisions an endpoint with default instance types and container images. Users can adjust instance size and region before confirming the deploy. The button shows up on models that already have SageMaker JumpStart support—roughly 15,000 models at launch.
What stands out
- 01No CLI required. The entire flow—from model card to live endpoint—happens in-browser. Users authenticate via AWS SSO and confirm instance settings in a modal.
- 02Pre-selected instance types. Hugging Face suggests a default SageMaker instance based on model size. A 7B parameter model typically lands on an
ml.g5.2xlarge; larger models getml.g5.12xlargeorml.p4d.24xlarge. - 03JumpStart catalog as the backend. Only models already listed in SageMaker JumpStart are eligible. That excludes some newer fine-tunes and community checkpoints that haven't been vetted for the catalog.
- 04Billing starts immediately. The endpoint runs on the user's AWS account and accrues charges from the moment it spins up. There's no free tier or trial period baked into the button.
