DeepSeek-R1 models now available on Amazon Bedrock and Amazon Sagemaker Jumpstart

You can now deploy DeepSeek-R1 models using Amazon Bedrock and Amazon SageMaker AI. Amazon Bedrock is ideal for teams looking to quickly integrate pre-trained foundation models via APIs, while Amazon SageMaker AI offers more flexibility for organizations that need advanced customization, training, and deployment with deeper access to infrastructure. Additionally, AWS Trainium and AWS Inferentia enable cost-effective deployment of DeepSeek-R1-Distill models via Amazon Elastic Compute Cloud (EC2) or Amazon SageMaker AI.

Using AWS, you can efficiently explore, create, and scale generative AI solutions while minimizing infrastructure expenses. The DeepSeek-R1 models offer a strong and economical approach to developing AI-driven applications, leveraging AWS’s solid security framework. For improved safety and reliability of your generative AI applications, it is advisable to integrate Amazon Bedrock Guardrails, which provide an extra layer of protection for users of both Amazon Bedrock and Amazon SageMaker AI.

There are multiple ways to deploy DeepSeek-R1 models on AWS, including:

1. Amazon Bedrock Marketplace for DeepSeek-R1 models
2. Amazon SageMaker JumpStart for DeepSeek-R1 models
3. Amazon Bedrock Custom Model Import for DeepSeek-R1-Distill models
4. Amazon EC2 Trn1 instances for DeepSeek-R1-Distill models

These deployment options allow you to choose the best approach to scaling and optimizing your generative AI applications on AWS.

Learn more on how to set up Deepseek-R1 on AmazonBedrock here and for Amazon Sagemaker Jumpstart here.