OpenAI lists Codex on Amazon Bedrock

OpenAI just gave enterprises a new way to buy and run its models. On April 28, OpenAI and AWS said OpenAI models, Codex, and Bedrock Managed Agents are now available through Amazon Bedrock. That matters because Bedrock is where a lot of large companies already handle model access, governance, billing, and security on AWS. So this is not just another model listing — it changes the route enterprises can use to get OpenAI inside production systems. (aws.amazon.com) ### What actually showed up on Bedrock? The new package is broader than a single API endpoint. AWS says Bedrock now offers OpenAI models, Codex, and Managed Agents powered by OpenAI. OpenAI frames Codex as its coding harness and product suite, and says organizations can now power Codex with OpenAI mod(aws.amazon.com)tepping outside their AWS environment. (aws.amazon.com) ### Why is Codex the interesting part? Because Codex is closer to “software that does work” than “model you prompt.” OpenAI describes it as a frontier coding harness, but the company also says customers are using it for research, analysis, and document-heavy work by connecting it to apps and tools the(aws.amazon.com)nterprise systems — code, documents, briefs, spreadsheets, and long-running tasks. (openai.com) ### Why does Bedrock change the buying motion? Bedrock is already the control plane for many AWS customers. AWS pitches the OpenAI offering as using the same Bedrock APIs, unified security, governance, and cost controls, with no extra infrastructure or new security model to learn. OpenAI adds the key procurement detail: eligible customers can apply Codex usag(openai.com)tence enterprise buyers care about more than benchmark charts. (aws.amazon.com) ### Does this mean OpenAI is no longer Azure-only? Basically, yes — at least in distribution terms. OpenAI has been widening its infrastructure and cloud relationships for months. In February, OpenAI and Amazon announced a strategic partnership around a Stateful Runtime Environment in Bedrock. In November 2025, OpenAI and AWS announced a multi-year partner(aws.amazon.com)capacity. This week’s Bedrock launch is the commercial surface area of that shift. (openai.com) ### What does AWS get out of it? A stronger answer to the “which model providers can I standardize on?” question. Bedrock has always been about model choice behind one API layer. Adding OpenAI’s flagship models and Codex makes that pitch much stronger, especially for companies that want OpenAI capability without changing cloud procurement, identity, or g(openai.com)the enterprise agent stack, not just raw infrastructure. (docs.aws.amazon.com) ### What does this change for developers? Less migration pain. AWS now supports OpenAI-compatible API endpoints for Bedrock models, and its docs say developers can use familiar OpenAI SDKs and tools with minimal code changes — basically swapping the base URL and API key. That lowers the friction for teams that built around OpenAI interfaces but need AWS-native controls. (docs.aws.amazon.com) ### Is there a catch? Yes — availability and branding are still a little fragmented. AWS says some of the latest OpenAI models are in preview, and Bedrock’s OpenAI pages emphasize enterprise controls more than consumer-style product simplicity. So the win here is not “the easiest way to use OpenAI.” It is “the easiest way for an AWS enterprise to govern OpenAI.” (aws.amazon.com) ### Bottom line? This is a distribution story dressed up as a product launch. OpenAI is meeting customers where their budgets, controls, and cloud commitments already live. And AWS just made Bedrock a much more credible home for enterprise teams that want OpenAI models — especially Codex — without reorganizing the rest of their stack. (aws.amazon.com)