AWS Brings OpenAI’s Open-Weight Models to Amazon Bedrock and SageMaker
This marks the first time OpenAI models are available through AWS.
Amazon Web Services (AWS) has announced that OpenAI’s open weight models—gpt-oss-120b and gpt-oss-20b—are now available via Amazon Bedrock and Amazon SageMaker, giving developers powerful tools to build generative AI applications at scale. This marks the first time OpenAI models are available through AWS.
OpenAI unveiled its two open-source models last night following months of anticipation. Within hours of their release, the models quickly became the top-trending models on Hugging Face.
Running natively in Amazon Bedrock, gpt-oss-120b is reported to be significantly more cost-effective than competing models—10x better than Gemini, 18x over DeepSeek-R1, and 7x over OpenAI’s own o4 model.
"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 at AWS.
These models are optimised for agentic AI workflows and include features like adjustable reasoning, tool use (e.g., web search, code interpreter), 128K context windows, and robust safety protocols. The launch adds to AWS’s expanding model roster, now boasting over 100 options from leading AI providers.
Recently, AWS laid off at least several hundred employees, following an internal strategic review. While the precise number remains undisclosed, the cuts impact multiple teams, including the customer-facing “specialists” group.
An Amazon spokesperson explained: “We’ve made the difficult business decision to eliminate some roles across particular teams in AWS,” adding the reductions are aimed at helping the company “continue to invest, hire, and optimize resources to deliver innovation for our customers.”
