Amazon AWS
Amazon Web Services (AWS) is the world's largest cloud provider. For LLMs it offers a fully managed model API (Amazon Bedrock), managed endpoints for deploying your own models (Amazon SageMaker AI) and GPU servers you run yourself (Amazon EC2).
Overview
Amazon Web Services (AWS) is Amazon's cloud computing business and the largest cloud provider in the world, with data centres in more than 30 regions across North America, South America, Europe, the Middle East, Africa and Asia Pacific.
AWS gives you three main ways to run large language models. Amazon Bedrock is a serverless API for ready-to-use models, including open-weight families like Llama, Mistral, DeepSeek, Qwen, Gemma and gpt-oss, billed per token. Amazon SageMaker AI lets you deploy open models, or your own fine-tuned versions, to managed endpoints billed per instance-hour. Amazon EC2 provides raw GPU and AI accelerator servers, from NVIDIA T4 up to B200, plus AWS's own Inferentia and Trainium chips, for teams that want full control.
AWS suits businesses that already run on AWS, need enterprise security and compliance, or want to keep data within a specific region.
Plans from Amazon AWS
| Hosting plan | Hosting type | Pricing range | Model families |
|---|---|---|---|
| Amazon Bedrock | Serverless API | US$0.04 – US$2.66 | Llama, DeepSeek, Qwen, Gemma, gpt-oss |
| Amazon SageMaker | Dedicated endpoint | From US$0.736 | Llama, Mistral, DeepSeek, Qwen, Gemma, gpt-oss |
| Amazon EC2 | GPU cloud | From US$0.526 | Mistral, DeepSeek, Gemma, gpt-oss, Llama, Qwen |