AWS made OpenAI GPT-6 Astra generally available on Amazon Bedrock as of September 14, 2026, allowing customers to run OpenAI’s latest model through the AWS service. The model supports a context window of up to 1 million input tokens, enabling users to submit large codebases, long contracts or extensive document collections in a single request.
AWS described GPT-6 Astra as OpenAI’s latest and most capable model to date, with deeper reasoning and judgment, professional-quality writing and design, and advanced computer and browser use for demanding business workflows. Those capability descriptions are AWS and OpenAI’s stated positioning for the model, rather than independently established results.
Customers can call GPT-6 Astra through supported Amazon Bedrock APIs or configure ChatGPT Work and Codex to use the model on Amazon Bedrock. OpenAI is also introducing enterprise plugins for ChatGPT Work that are intended to extend Astra’s browser-use capabilities across common business applications.
According to AWS, established controls can be used to secure workloads, govern access and audit model invocation activity. AWS also states that customers’ inference data is not used for model training.
The model release led AWS’s weekly product update, which also included general availability of the Amazon Quick desktop application for macOS and Windows. AWS said the application can work with local files, remain connected to calendars, email and business applications, and synchronize conversations, context and agents across desktop and mobile devices.
Other announced changes included a 90-minute timeout for asynchronous and event source mapping invocations on Lambda Managed Instances, six times the previous 15-minute limit, while synchronous invocations retain the 15-minute maximum. AWS also announced cross-account copying for Amazon EBS Volume Clones, general availability of second-generation single-rack AWS Outposts, Amazon OpenSearch Serverless support in v0 by Vercel, and general availability of AWS Transform for .NET modernization through its command-line interface.