AWS Integrates OpenAI’s First Open-Source AI Models
-
Amazon Web Services (AWS) has integrated OpenAI’s first open-weight models into its cloud infrastructure, making the company’s technology available through AWS’s platform for the first time. The cloud computing division of Amazon has added two new foundation models from OpenAI – gpt-oss-120b and gpt-oss-20b – to its Amazon Bedrock and Amazon SageMaker AI services. Amazon says this technology enables 'customers to quickly and easily build Gen AI applications.' OpenAI’s two new open weight foundation models will put more powerful AI technologies into the hands of organisations and expand the impact of OpenAI’s leading technology by making it available to the millions of customers on AWS. Open-weight models allow developers to access and modify the underlying parameters of AI systems, unlike closed models where the internal workings remain proprietary. Internal benchmarking suggests the OpenAI models operate at 10 times better price-performance than comparable models from Google’s Gemini, 18 times better than DeepSeek-R1 and seven times better than OpenAI’s own o4 model when running on Amazon Bedrock. The integration focuses particularly on AI agents – software systems that can perform complex tasks autonomously. These agents are a growing area of enterprise AI deployment, where systems handle multi-step processes without human intervention. 'Open weight models are an important area of innovation in the future development of Gen AI technology, which is why we have invested in making AWS the best place to run them – including those launching today from OpenAI,' says Atul Deo, Director of Product at AWS. Amazon Bedrock, AWS’s managed AI service, already serves tens of thousands of customers across industries including startups and Fortune 500 companies. The service provides enterprise-grade security features and tools such as Guardrails, which AWS says blocks up to 88% of harmful content through configurable safety measures. Now with OpenAI, its models support instruction-following and tool use capabilities including web search and code interpretation. These features enable the models to access real-time information and execute multi-step tasks, making them suitable for applications in coding, scientific analysis and mathematical problem-solving. The models feature a 128,000-token context window, allowing processing of lengthy documents such as customer service transcripts, technical documentation and academic papers. This context length enables the models to maintain coherence across extended conversations and complex document analysis. 'Our open weight models help developers – from solo builders to large enterprise teams – unlock new possibilities across industries and use cases,' says Dmitry Pimenov, Product Lead at OpenAI. 'Together with AWS, we’re providing powerful, flexible tools that make it easier than ever for customers to build, innovate and scale.' OpenAI has implemented safety training and evaluation for both models, addressing concerns about responsible deployment of AI systems. The models include adjustable reasoning levels and chain-of-thought outputs, which break down complex problems into logical steps for transparency. AWS customers including DoorDash, the food delivery service, GoDaddy, the web hosting company and pharmaceutical giant Pfizer already use Amazon Bedrock for various AI applications. The addition of OpenAI models expands the range of available technologies alongside existing providers including Meta, Mistral AI and DeepSeek. Amazon SageMaker AI, AWS’s machine learning (ML) platform, will also provide access to the OpenAI models alongside tools for pre-training, evaluation, fine-tuning and deployment. This integration allows customers to customise the models for specific use cases while maintaining enterprise security requirements. More broadly, the partnership is AWS’s strategy of offering multiple AI model providers rather than developing proprietary alternatives. Amazon Bedrock currently hosts more than 100 models from various AI companies, positioning AWS as a platform provider rather than a direct competitor to model developers. 'The addition of OpenAI as our newest open weight model provider marks a natural progression in our commitment to bringing cutting-edge AI to organisations worldwide – and the unmatched size of our customer base marks a transformative shift in access to OpenAI’s advanced technology,' says Atul. 'Model selection and customer choice matters,' says Jon Jones, Vice President of Global Startups and Venture Capital at AWS.