Amazon Web Services (AWS) announced on August 5 that OpenAI’s new open weight foundation models—gpt-oss-120b and gpt-oss-20b are now available on Amazon Bedrock and Amazon SageMaker AI, marking a significant expansion in generative AI offerings on its cloud platform.
This is the first time OpenAI’s open weight models are being made accessible through AWS, offering customers a streamlined path to build and deploy generative AI applications using familiar AWS tools and infrastructure.
The larger model, gpt-oss-120b, delivers 3x better price-performance than Google’s Gemini, 5x better than DeepSeek-R1, and 2x better than OpenAI’s own o4 model, according to Amazon.
With advanced reasoning abilities, these models are particularly suited for agentic workflows, programming, scientific analysis, and multi-step problem-solving. They are equipped with features such as chain-of-thought outputs, web search, and code interpretation, alongside a 128K token context window to process large documents.
OpenAI models come integrated with Amazon Bedrock's Guardrails, preventing up to 88% of toxic content with configurable filters. The models also passed thorough safety evaluation to enable responsible deployment.
On Amazon SageMaker AI, customers are able to use OpenAI models for pre-training, evaluation, fine-tuning, and custom deployment, using AWS's strong set of machine learning tools.
Amazon announced that companies like Pfizer, Siemens, the PGA TOUR, and GoDaddy already leverage Amazon Bedrock for GenAI-driven transformations. OpenAI now joins Meta, Mistral AI, and others as part of AWS's expanding ecosystem of open weight model providers, it announced.
