AWS Launches OpenAI Models & Agentic AI Tools: Bedrock, Quick & Connect Updates

Amazon Web Services (AWS) on Tuesday launched a series of initiatives designed to bring OpenAI’s most powerful models to its Bedrock platform, unveil a new agentic developer framework, release a desktop AI productivity tool called Amazon Quick, and expand its Amazon Connect service. The announcements, made at the “What’s Next with AWS” event in San Francisco, followed a restructuring of the exclusive cloud partnership between OpenAI and Microsoft, freeing OpenAI to distribute its products across rival cloud providers.

AWS CEO Matt Garman described the partnership with OpenAI as “huge,” noting that customers have been requesting access to OpenAI models within AWS “from the very early days.” Amazon CEO Andy Jassy had previously signaled his anticipation of these developments, calling the Microsoft-OpenAI restructuring “very interesting” in a post on X the day prior.

OpenAI’s Models Arrive on Amazon Bedrock

The centerpiece of the announcement is the availability of OpenAI’s latest models through Amazon Bedrock in limited preview, with general availability expected within weeks. AWS confirmed that GPT-5.4 is available immediately in limited preview, with GPT-5.5 arriving shortly thereafter. This integration allows AWS customers to evaluate and deploy OpenAI models alongside offerings from Anthropic, Meta, Mistral, Cohere, and Amazon’s own models—all through Bedrock’s unified security, governance, and cost controls.

Anthony Liguori, Vice President and Distinguished Engineer at AWS, emphasized the significance of this moment. “Customers can take their existing workloads today and just start using AWS right off the bat,” he said. “They don’t have to write any new software, develop any new things. I think that’s one of the most exciting announcements that came out today.”

A Shift in the Cloud AI Marketplace

The path to Tuesday’s announcement was complicated by a previous $50 billion deal between OpenAI and Amazon, which created a legal tangle with Microsoft. Under the original Microsoft-OpenAI agreement, Microsoft retained exclusive rights to OpenAI products accessed through APIs, potentially conflicting with OpenAI’s promise to give AWS exclusive hosting rights for its new Frontier agent-building tool. Microsoft initially pushed back, stating that “Azure remains the exclusive cloud provider of stateless OpenAI APIs.” However, Monday’s restructured deal—replacing Microsoft’s open-ended exclusivity with a nonexclusive license running through 2032—resolved these legal obstacles.

Denise Dresser, OpenAI’s revenue chief, explained that the Microsoft relationship “has also limited our ability to meet enterprises where they are — for many that’s Bedrock.” She framed the moment as a turning point, stating that enterprise customers are “no longer in the mindset of experimentation and pilots” and want to deploy AI “full enterprise wide” in a “trusted environment.”

“We are co-developing an agent platform from the ground up, deeply integrated with AWS services and powered by OpenAI’s most advanced models and tools,”

Sam Altman, OpenAI CEO

Altman, who was unable to attend in person due to a court case, stated that OpenAI is co-developing an agent platform deeply integrated with AWS services and powered by its most advanced models and tools, allowing customers to build and run powerful agents in their own environment.

Bedrock Managed Agents and the Reinforcement Learning Approach

AWS also launched Amazon Bedrock Managed Agents powered by OpenAI, combining OpenAI’s frontier models with its proprietary “harness,” the agentic execution framework that powers Codex. Liguori explained that the harness concept represents a shift in how models are trained and deployed for agentic work. He argued that the best agentic performance comes when models are trained specifically against their harness through reinforcement learning.

“You can give a model a whole lot of instructions and a set of tools, and it will be able to use it most of the time,” he said. “But when you really train the model on a specific set of tools, a specific style of operations, it’s just like drilling plays over and over again — the model builds muscle memory for using that harness.”

Anthony Liguori, Vice President and Distinguished Engineer at AWS

Bedrock Managed Agents consists of a runtime layer for configuring skills, memory policies, and tool access; an environment layer for agent deployment; and an inference API for interaction. The system integrates with AWS’s identity and access management, VPC networking, and CloudTrail auditing.

Security and Infrastructure

Liguori made a significant claim regarding security, stating that the system hosting the GPT-5.4 models is “zero operator access,” meaning no human can log into the machines running inference data. This security is built on AWS’s custom silicon—Graviton processors and Nitro security chips—with only memory modules sourced off-the-shelf.

Codex and Amazon Quick

OpenAI’s coding agent, Codex, is now available on Bedrock in limited preview. Dresser shared that Codex has grown from 3 million to 4 million weekly active users in two weeks. AWS also launched Amazon Quick Desktop, a desktop application designed to bring agentic AI to knowledge workers, integrating with Google Workspace, Microsoft 365, Zoom, and Salesforce.

Amazon Connect Expands

Amazon Connect is expanding from a single contact-center product into a family of four agentic AI solutions: Amazon Connect Decisions for supply chain planning, Amazon Connect Talent for hiring, and Amazon Connect Health for healthcare. Colleen Aubrey, who leads applied AI solutions at AWS, introduced the concept of “humorphism”—translating human interaction dynamics into AI agent behavior—as a guiding design principle for these new solutions.

These launches represent AWS’s bid to become the definitive infrastructure layer for the agentic AI era, with a four-layer strategy encompassing custom infrastructure, model access, an agentic platform, and purpose-built applications. The company believes this approach will capture value across the entire stack and address the evolving needs of enterprises in the AI landscape.