Deploy DeepSeek-R1 on Amazon Bedrock: A Complete Beginner’s Guide

Amazon Bedrock Marketplace is a new capability in Amazon Bedrock that enables developers to discover, test and use over 100 popular, emerging, and specialized foundation models (FMs) alongside the current selection of industry leading models in Amazon Bedrock. You can easily discover the models in a single catalog, subscribe to the model and then deploy the model on managed endpoints. Then you can access the models through Bedrock’s unified APIs and use them natively with Bedrock tools such as Agents, Knowledge Bases and Guardrails.

In this guide, we will walk through how to set up DeepSeek-R1 on AWS Bedrock, configure necessary permissions, test model inference and optimize for cost-effective usage.

• Fully Managed LLM Deployment: No need to set up servers, GPUs, or infrastructure.
• Scalability: Automatically scales based on demand.
• Pay-As-You-Go: Optimized pricing, paying only for what you use.
• Security & Compliance: AWS IAM-based access control, data encryption.
• Bedrock API Integration: Easily integrates with AWS services (S3, Lambda, SageMaker, etc.).

• AWS Account (With access to Bedrock Marketplace)

• IAM Permissions (AmazonBedrockFullAccess, AWSLambdaFullAccess)

• S3 Bucket (For logging and data storage)

• AWS SDK or CLI (For API calls)

Setting Up IAM Role for Bedrock

1. Go to AWS IAM Console → Click Roles → Create role

2. Select AWS Service → Choose Bedrock

3. Attach Policies:

• AmazonBedrockFullAccess

• AWSLambdaFullAccess

4. Review & Create Role (Name it BedrockAccessRole)

Since AWS Bedrock operates on a pay-per-inference model, it’s crucial to optimize usage.

Pricing Based on Instance Type

1. ml.g5.2xlarge ($0.75/hr) – Best for testing & development.
2. ml.p4d.24xlarge ($32.00/hr) – Ideal for high-performance workloads.
3. ml.trn1.32xlarge ($25.00/hr) – Cost-effective for large-scale processing.

How to Optimize Costs?

• Use Auto-Scaling to avoid over provisioning
• Deploy in a Cheaper AWS Region (e.g., Ohio instead of N. Virginia)
• Batch Requests to maximize model usage per API call

Step 1: Go to Amazon Bedrock Console

• Navigate to Model Catalog under the Foundation Models section.

• Search for "DeepSeek-R1" or filter by model providers.

Step 2: Review Model Details

• Check the model's capabilities, implementation guidelines and use cases.

Step 3: Deploy the Model

• Provide an endpoint name.
• Choose the number of instances and instance type.
• Configure advanced security and infrastructure settings (e.g., VPC networking, encryption, service role permissions).


 

Ensure settings align with security and compliance requirements for production use.

Using Amazon Bedrock Guardrails

• Bedrock Guardrails help control user interactions with DeepSeek-R1 by filtering harmful content.
• The model can only be used with Bedrock’s ApplyGuardrail API for evaluating inputs and responses.
• Guardrails integrate with Bedrock Agents and Bedrock Knowledge Bases for safer AI applications.

Testing and Optimizing DeepSeek-R1

• Use the Bedrock playground to test how the model responds to different inputs.
• Fine-tune prompts for better results.

When using InvokeModel API or Bedrock’s playground, use DeepSeek’s chat template for best performance:

Ready to transform your business with our technology solutions? Contact Us  today to Leverage Our AI/ML Expertise.