OpenAI’s open weight models available today in Amazon Bedrock and Amazon SageMaker AI

Key takeaways

AWS will offer open weight models from OpenAI, making their models available to the millions of customers on AWS for the first time.
Customers can leverage the new open weight models’ advancing reasoning capabilities for agentic workflows, coding, scientific analysis, and mathematical problem-solving.
This news expands the selection of models on AWS, empowering customers with open weight models, which will help shape the future of GenAI technology.

Today, Amazon Web Services (AWS) announced that OpenAI open weight models will be available via Amazon Bedrock and Amazon SageMaker AI for the first time, enabling customers to quickly and easily build generative artificial intelligence (AI) applications. OpenAI’s two new open weight foundation models—gpt-oss-120b and gpt-oss-20b—will put more powerful AI technologies into the hands of organizations and expand the impact of OpenAI’s leading technology by making it available to the millions of customers on AWS, the world’s most comprehensive and broadly adopted cloud. Running in Amazon Bedrock, the larger of the two open weight models is 3x more price-performant than the comparable Gemini model, 5x more than DeepSeek-R1, and 2x more than the comparable OpenAI o4 model. This news continues AWS’s commitment to model choice by enhancing the existing broad selection of fully managed models in Amazon Bedrock, as well as by making them accessible in Amazon SageMaker JumpStart, and reinforces AWS’s momentum in delivering innovative generative AI technologies to meet customers’ needs.

The advanced reasoning capabilities of OpenAI’s open weight models make them well suited for use cases involving AI agents, which are transforming how organizations conduct business. Using Amazon Bedrock AgentCore, customers can deploy and operate highly effective agents using OpenAI models running natively on Amazon Bedrock to provide them with the scale and security required for production applications. Customers can also seamlessly integrate gpt-oss-120b and gpt-oss-20b with Amazon Bedrock’s enterprise-grade security and powerful tools like Guardrails—which helps block up to 88% of harmful content using configurable safeguards—with support for Custom Model Import, Knowledge Bases, and customization capabilities coming in the future. With Amazon SageMaker AI, customers can leverage OpenAI's open weight models alongside comprehensive tools for pre-training, evaluation, fine-tuning, and model deployment.

"Open weight models are an important area of innovation in the future development of generative AI technology, which is why we have invested in making AWS the best place to run them—including those launching today from OpenAI," said Atul Deo, director of product, AWS. "The addition of OpenAI as our newest open weight model provider marks a natural progression in our commitment to bringing cutting-edge AI to organizations worldwide, and the unmatched size of our customer base marks a transformative shift in access to OpenAI’s advanced technology."

"Our open weight models help developers—from solo builders to large enterprise teams—unlock new possibilities across industries and use cases," said Dmitry Pimenov, product lead, OpenAI. "Together with AWS, we’re providing powerful, flexible tools that make it easier than ever for customers to build, innovate, and scale."

As organizations increasingly embrace agentic AI, they need a variety of models capable of executing complex tasks. OpenAI’s open weight models meet this need with a compelling performance-to-size ratio and advanced reasoning, including adjustable reasoning levels and chain-of-thought outputs that break down complex problems into logical steps. This makes them great for use cases like agentic workflows, coding, scientific analysis, and mathematical problem-solving. These models support instruction-following and tool use—web search and code interpreter—to help them reference real-time information and perform multi-step tasks. Featuring a 128K context input window, these text-generation models enable customers to process longer documents and conversations, such as customer service transcripts, detailed technical documentation, academic papers, and more. Safety is a central component of the open weight models, and to support the responsible deployment of generative AI applications, each model has undergone comprehensive safety training and evaluation by OpenAI.

Amazon Bedrock is already empowering tens of thousands of global customers of all sizes and industries—from fast-growing startups to established Fortune 500 companies and trusted government organizations—with the flexibility to securely and responsibly build, customize, and innovate on generative AI applications as their needs evolve. Customers including DoorDash, GoDaddy, Lonely Planet, LexisNexis Legal & Professional, Pfizer, the PGA TOUR, Rocket Companies, Siemens, and TUI are using Amazon Bedrock to modernize business processes, develop next-generation applications, and transform their enterprises. OpenAI joins other open weight model providers such as DeepSeek, Meta, Mistral AI, and others that have made their technologies available via Amazon Bedrock and Amazon SageMaker AI, further expanding customers' ability to select and combine the optimal models for their unique requirements. Today’s news also further enhances Amazon Bedrock’s existing selection of more than 100 models from leading AI companies, including fully managed models from providers like Luma AI, TwelveLabs, and Writer that are made available on AWS as the only cloud service provider to date.

OpenAI open weight models available today on AWS

Key takeaways

AWS announces new innovations for building AI agents at AWS Summit New York 2025

How Amazon is helping power America's AI future