OpenAI Brings GPT-5.5, Codex, and Managed Agents to AWS Bedrock

OpenAI has put its frontier models, its coding product suite, and a new managed agent service on Amazon Bedrock, marking a major expansion of its partnership with AWS and a sharp shift in how the company sells to enterprises. All three offerings — OpenAI models on AWS, Codex on AWS, and Amazon Bedrock Managed Agents, powered by OpenAI — launch today in limited preview.

The headliner is GPT-5.5, which OpenAI calls its "best frontier model," now available directly through Amazon Bedrock. The move gives AWS customers a way to run OpenAI models alongside the services, security controls, identity systems, and procurement processes they already rely on, rather than wiring OpenAI's own API into environments governed by AWS policy.

The stakes are straightforward. Large enterprises increasingly buy AI through their existing cloud commitments, and until now OpenAI's models were largely absent from the two clouds that dominate that procurement path — AWS and, via a separate arrangement, Google Cloud. Putting GPT-5.5 on Bedrock puts OpenAI inside the billing, governance, and compliance perimeter where most enterprise workloads already run, positioning it directly against Anthropic, whose Claude models have been a flagship Bedrock offering.

For developers, OpenAI frames the benefit as flexibility: new AI applications, intelligence embedded in existing products, or agentic workflows that "can reason, take action, and support more complex business processes." For enterprises, the company pitches "a clear single path from experimentation to production, with OpenAI capabilities available in the AWS environments where their most important workloads already run."

Codex comes to Bedrock

The second piece of the announcement is Codex, OpenAI's frontier coding harness and product suite. More than 4 million people now use Codex every week, according to OpenAI, and usage has spread well beyond writing code. Teams use it across the software development lifecycle — to write code, explain systems, refactor applications, generate tests, and modernize legacy codebases — and increasingly to accelerate research, analysis, and document-based work, from summarizing source materials to creating briefs, slide decks, and spreadsheets.

Organizations can now power Codex with OpenAI models served directly from Amazon Bedrock. That means any company with an AWS commit and Bedrock access can start using OpenAI's coding agent without a separate vendor relationship. Customers get started by configuring Codex to use Bedrock as the provider, which OpenAI says delivers the enterprise-grade attributes they expect from AWS — security, billing, and high availability.

All customer data is processed by Amazon Bedrock, and eligible customers can apply Codex usage toward their AWS cloud commitments — a detail that matters for enterprises negotiating multi-year spend agreements with Amazon. Codex on Bedrock is available in limited preview, with configuration available through the Bedrock API starting with Codex CLI, the Codex desktop app, and the Visual Studio Code extension.

Managed Agents handle the hard parts

The third component is the newest: Amazon Bedrock Managed Agents, powered by OpenAI. The service gives enterprises a way to deploy advanced agents within their existing AWS environments, with the agents capable of maintaining context, executing multi-step workflows, using tools, and taking action across complex business processes.

The pitch is that teams can focus on making agents useful for real work rather than assembling the infrastructure around them. Managed Agents handles deployment, tool use, orchestration, and governance, with built-in integration across Amazon's security and compliance controls. OpenAI says the result is "a faster path from prototype to production for agents that can operate in real enterprise environments."

That framing lands in a crowded field. Agent deployment has become the defining enterprise AI battleground of the past year, with vendors competing on orchestration, tool use, and governance rather than raw model quality. A managed service that folds OpenAI's models into AWS's compliance stack removes one of the main friction points — infrastructure assembly — that has kept many agent projects stuck at the prototype stage.

Why it matters

Taken together, the three launches signal that OpenAI is treating distribution through hyperscale clouds as a first-class channel, not an afterthought. Enterprise AI adoption has consistently bottlenecked on procurement, security review, and data residency — not model capability. By running inside Bedrock, OpenAI inherits AWS's answers to all three.

The announcement also deepens the OpenAI–Amazon strategic partnership at a moment when model providers are racing to lock in enterprise relationships through whichever cloud a customer already uses. For AWS, adding GPT-5.5 broadens Bedrock's model portfolio; for OpenAI, it puts frontier models in front of customers who would otherwise default to incumbent Bedrock providers.

Everything announced today ships in limited preview, so broad availability and pricing details remain open questions. But the direction is clear: OpenAI wants its models, coding tools, and agents to run wherever enterprise workloads already live — and it is now willing to meet customers inside AWS to make that happen.