Published: June 26, 2026

OpenAI Models Now Run Inside AWS — Breaking Their Exclusivity with Microsoft

Modelos de OpenAI integrados en AWS Bedrock

Since June 1st, OpenAI's models are available within Amazon Bedrock, AWS's AI platform. Until now, a company wanting to use GPT securely on AWS had to set it up separately. Now, both are integrated into the same environment. An important note: we are talking about models via API — GPT-5.5 and GPT-5.4 — and the Codex code agent, not the ChatGPT app.

What Has Been Integrated

GPT-5.5, OpenAI's most capable model, and GPT-5.4, its best value-for-money option, both accessible through the Responses API. Additionally, Codex, OpenAI's programming agent used by over 4 million people weekly. These join the Bedrock catalog alongside Anthropic, Meta, Mistral, and others — all managed from a single console, with the same security controls, governance, and AWS cost management.

What Benefits Does a Company Gain

- Security and governance: each call inherits the controls you already use on AWS — identity, network isolation, encryption, audit logging. Your data is not used for training.

- No need to change environments: evaluate and deploy GPT alongside other models without a new provider, invoice, or process.

- No additional costs: pricing is the same as directly contracting OpenAI, and usage counts toward your AWS spending commitments.

Important Information to Know

Currently available in limited U.S. regions (Ohio, Oregon, Virginia, and GovCloud) with a context window of 272,000 tokens in Bedrock. If your use case depends on a specific region or feature, test it before committing to deployment.

Our Perspective

Until recently, OpenAI operated exclusively on Microsoft's cloud. The fact that its models are now available on AWS — where Claude already resides — means that model selection is becoming independent of cloud provider choice: you select the model based on how it fits your work, not which cloud you're tied to. For anyone building AI-based products, this means more options and less dependency on a single provider.

Read more news

More recent articles to keep exploring the blog.