How to deploy and fine-tune DeepSeek models on AWS
Hugging Face Blog
Hugging Face and AWS published documentation for deploying and fine-tuning DeepSeek-R1 models across AWS services including SageMaker, Inference Endpoints, and Bedrock. Deployment on Hugging Face Inference Endpoints costs 8.3 dollars per hour, while the distilled models can run on various GPU instances ranging from ml.g6.2xlarge to ml.g6.48xlarge depending on model size. Developers can now deploy open-source reasoning models with multiple configuration options, though fine-tuning support is still in development.
.png)
