September 9, 2026, (Inside AI) — OpenAI and Amazon Web Services have made GPT-6 Astra generally available on Amazon Bedrock, the companies announced. The model, described as OpenAI's most capable to date, targets enterprise workflows requiring deep reasoning, professional writing, and browser-based automation.
The launch places GPT-6 Astra directly inside AWS's managed AI platform, letting developers call the model through supported Bedrock APIs or configure ChatGPT Work and Codex to run on Bedrock. It supports up to 1 million input tokens and can align output with a company's voice, templates, and standards.
This move deepens the commercial relationship between two firms often seen as rivals in the AI infrastructure race. AWS has long positioned Bedrock as a neutral marketplace for frontier models. Adding OpenAI's flagship model signals that enterprise demand for choice now outweighs competitive posturing.
The general availability comes after months of limited previews. AWS said the Bedrock inference engine provides the performance, security, and scale needed for production workloads. Enterprises can govern access, audit model invocation, and secure workloads using established AWS controls.
OpenAI is also introducing new enterprise plugins for ChatGPT Work. These plugins extend Astra's browser-use capabilities across common business applications. The goal is to let teams automate multi-step tasks inside tools they already use, without custom integrations.
Use cases cited include building autonomous agents, analyzing large document collections, investigating complex software issues, and creating applications that weigh competing inputs. The 1 million token context window is a key differentiator. It lets the model process entire codebases, lengthy contracts, or research archives in a single call.
Pricing and regional availability vary. AWS pointed users to the Amazon Bedrock documentation for supported regions, endpoints, APIs, features, inference profiles, and pricing. A companion blog post outlines what developers can build with GPT-6 Astra.
Why Astra's Bedrock Debut Signals a Shift in Enterprise AI
For years, enterprises chose between OpenAI's direct API and cloud marketplaces. Bedrock's value proposition is operational simplicity: one API, unified billing, and built-in governance. By making GPT-6 Astra available there, OpenAI acknowledges that many large customers prefer to consume AI through their existing cloud provider.
This is not OpenAI's first appearance on a major cloud. Microsoft Azure has long hosted OpenAI models. But AWS is the largest cloud infrastructure provider. Making Astra available on Bedrock expands OpenAI's reach to thousands of enterprises that standardized on AWS for security and compliance reasons.
The browser-use capability is particularly notable. Astra can navigate web applications, fill forms, extract data, and complete multi-step tasks. Enterprise plugins for ChatGPT Work extend this to tools like CRM systems, ERP platforms, and internal dashboards. That moves the model from a text generator to an action-taking agent.
Industry analysts have noted that agentic AI is the next battleground. Models that can reason, plan, and execute tasks across software systems command premium pricing. Astra's positioning on Bedrock suggests OpenAI wants to capture that spend before competitors like Anthropic or Google solidify their enterprise footholds.
What Enterprise Buyers Should Watch
Despite the fanfare, several questions remain. Pricing for a 1 million token context window can be substantial, especially for high-volume workloads. AWS has not disclosed specific per-token rates in the announcement, instead directing users to documentation. Enterprises should model costs carefully before migrating from smaller models.
Data governance is another consideration. While AWS controls help audit invocation, organizations must still evaluate where inference occurs and how data flows between OpenAI and AWS systems. For regulated industries, these details matter as much as model capability.
OpenAI's enterprise plugin ecosystem is young. The new plugins for ChatGPT Work are promising but unproven at scale. Buyers should test browser automation against real business applications before committing to production deployments. Early adopters may encounter integration friction that documentation does not yet capture.
Still, the general availability of GPT-6 Astra on Bedrock is a milestone. It brings one of the most advanced reasoning models into the largest managed AI marketplace. For enterprises already on AWS, the barrier to experimentation just dropped significantly.
