What happens when I click 'Customize on SageMaker AI' or 'Deploy on SageMaker AI'?

You are prompted to sign in to AWS (or skip if you have an active session), then land directly on either the Model Customization page or the Deployment page inside SageMaker Studio with your selected model pre-loaded. A new domain is provisioned automatically with pre-configured permissions.

Do I need to manually set up AWS permissions and GPU quotas?

No. A new managed policy, AmazonSageMakerModelCustomizationCoreAccess, is created and attached automatically. The Studio UI also now shows GPU quota availability (G5, G6 instance types) directly in the instance selection list, so you can see limits immediately without navigating to Service Quotas.

What fine-tuning methods are supported?

The pre-configured permissions support supervised fine-tuning (SFT), direct preference optimization (DPO), reinforcement learning with verifiable rewards (RLVR), and reinforcement learning from AI feedback (RLAIF), with deployment to SageMaker AI or Amazon Bedrock endpoints.

Open-Source AIHugging Face BlogJul 8, 2026Jul 8, 2026, 10:00 JST2 min read

Hugging Face, Amazon SageMaker link up for one-click model deployment

Send on LINE

Key takeaway

Hugging Face and Amazon SageMaker AI now integrate so developers can click from a model page directly into Studio with the environment fully configured and ready to fine-tune or deploy.
Previously, this workflow required multiple manual setup steps.
The integration is live today on supported models.

3 Key Points

What happened
Hugging Face and Amazon SageMaker AI announced a deep-link integration that lets developers move directly from discovering a model on Hugging Face to experimenting in SageMaker Studio. When clicking "Customize on SageMaker AI" or "Deploy on SageMaker AI" on supported models, developers land in Studio with the model pre-loaded, permissions pre-configured, and GPU quota visibility built in.
Why it matters
Previously, setting up SageMaker Studio after finding a model on Hugging Face required navigating multiple steps—opening the AWS Console, creating a domain, configuring permissions, sometimes requesting GPU quota—which slowed the path from discovery to hands-on work. The integration removes that friction, letting developers iterate faster without manual environment setup or permission troubleshooting. For teams using both platforms, this means staying focused on the task rather than switching between tools.
What to watch
The integration is available today on supported Hugging Face models. A new managed AWS policy, AmazonSageMakerModelCustomizationCoreAccess, is automatically created and attached, providing permissions for fine-tuning methods including supervised fine-tuning (SFT), direct preference optimization (DPO), reinforcement learning with verifiable rewards (RLVR), and reinforcement learning from AI feedback (RLAIF).

Ask the AI about this article →

Summaries like this, in your inbox every morning.

Get news like this every morning — free

Context & Analysis

The integration addresses a longstanding pain point in the model-to-deployment workflow. Developers using Hugging Face as a discovery platform and SageMaker as an experiment and deployment platform previously had to manually recreate context and permissions across both systems—a friction point that slowed iteration. By automating domain provisioning, attaching permissions on first use, and showing GPU quota availability upfront, Hugging Face and AWS have removed three of the most common setup bottlenecks.

The timing reflects a broader shift in how model development works. Developers now routinely browse pre-trained models, customize them on their data, and deploy them to managed infrastructure, often all in the same session. Hugging Face sits at the discovery layer, while SageMaker handles training and deployment. The deep-link integration makes both platforms feel like a single workflow rather than two separate services, reducing cognitive load and tool-switching overhead.

For AWS, the move also tightens Hugging Face's integration with its cloud ecosystem—potentially lowering friction for teams deciding where to host fine-tuning and inference workloads. For developers, the value is simpler: staying in flow instead of navigating menus and configuration dialogs.

FAQ

What happens when I click 'Customize on SageMaker AI' or 'Deploy on SageMaker AI'?: You are prompted to sign in to AWS (or skip if you have an active session), then land directly on either the Model Customization page or the Deployment page inside SageMaker Studio with your selected model pre-loaded. A new domain is provisioned automatically with pre-configured permissions.
Do I need to manually set up AWS permissions and GPU quotas?: No. A new managed policy, AmazonSageMakerModelCustomizationCoreAccess, is created and attached automatically. The Studio UI also now shows GPU quota availability (G5, G6 instance types) directly in the instance selection list, so you can see limits immediately without navigating to Service Quotas.
What fine-tuning methods are supported?: The pre-configured permissions support supervised fine-tuning (SFT), direct preference optimization (DPO), reinforcement learning with verifiable rewards (RLVR), and reinforcement learning from AI feedback (RLAIF), with deployment to SageMaker AI or Amazon Bedrock endpoints.

Hugging Face BlogRead Original Article

Send on LINE

Get the latest Open-Source AI news every morning

For example, today's edition would include:

DataAgent launches with $10M to auto-fix Kubernetes faultsSiliconANGLE AI · 4h ago
Z.ai runs GLM on 100,000 Chinese AI chipsDIGITIMES Asia · 7h ago
Broadcom Unveils VMware AI Factory for Faster Private AITop Companies AI · 16h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Subscribe free →

Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →