First, the news. On April 28, 2026, AWS launched OpenAI flagship models on Amazon Bedrock. Furthermore, this came one day after Microsoft seven-year exclusivity deal with OpenAI ended. So the GPT-5.5 AWS Bedrock integration is one of the biggest cloud announcements of the year.
If you have waited for OpenAI on AWS, the wait is over. Therefore, here is what shipped, what it means, and how to start without rebuilding your stack.
Key Takeaways
- OpenAI on AWS — Furthermore, GPT-5.5 and GPT-5.4 are in limited preview on Amazon Bedrock as of April 28, 2026.
- End of exclusivity — In addition, the launch came one day after Microsoft seven-year deal ended.
- Native AWS integration — Moreover, the models inherit IAM, PrivateLink, guardrails, encryption, and CloudTrail.
- Cost controls — As a result, token spend counts toward AWS EDPs and Savings Plans.
- Codex and agents too — However, OpenAI Codex and Managed Agents are also in limited preview on Bedrock.
Table of Contents
- What Is GPT-5.5 on AWS Bedrock?
- Why the GPT-5.5 AWS Bedrock Launch Matters
- GPT-5.5 AWS Bedrock Features
- Pricing and Availability
- How to Start Using It
- GPT-5.5 AWS Bedrock Use Cases
- vs OpenAI Direct API
What Is GPT-5.5 on AWS Bedrock?
So GPT-5.5 on Amazon Bedrock is OpenAI frontier model running inside the AWS managed AI service. The integration lets you call GPT-5.5 through the same Bedrock APIs. Therefore, no new SDK and no separate billing relationship.
Amazon Bedrock is AWS managed service for foundation models. Until April 2026, it shipped Anthropic, Mistral, Meta, Cohere, AI21, Stability AI, and Amazon. However, no OpenAI was on the list. As a result of the GPT-5.5 AWS Bedrock release, that gap closed.
Sound familiar? In contrast to before, you can now swap GPT models without re-architecting security.
The launch also matters for AWS competitive position. Furthermore, until now Azure was the only major cloud with OpenAI as a first-class service. So with GPT-5.5 AWS Bedrock live, AWS customers stay inside their preferred cloud.
Why the GPT-5.5 AWS Bedrock Launch Matters
For seven years, OpenAI frontier models lived only on Azure. Therefore, AWS-first teams had to build twice, secure twice, and audit twice. As a result, the GPT-5.5 AWS Bedrock launch removes that friction.
The competitive shift is just as big. According to Fortune, AI-related capex across Microsoft, Meta, Google, and Amazon will pass $600 billion in 2026. Furthermore, Alphabet alone raised its full-year guidance to $180 to $190 billion. So AWS landing OpenAI puts it back in direct contention for AI workloads.
In my experience with AWS-first teams, running a second cloud just for OpenAI was a real cost. As a result, removing that overhead is the practical story behind the headline.
The launch also lands at a key time for procurement. Furthermore, public-sector buyers in the UK and EU often need cloud services through pre-approved frameworks. Therefore, GPT-5.5 AWS Bedrock fits that need.
For more on managing cloud spend, see our cloud cost optimization guide.
GPT-5.5 AWS Bedrock Features
So the headline feature is parity. The model on Bedrock looks and feels like every other Bedrock model.
According to the AWS Bedrock OpenAI page, the integration ships with these by default:
- IAM integration — Furthermore, every model call uses your existing IAM policies.
- PrivateLink — In addition, traffic stays inside your VPC.
- Bedrock Guardrails — Moreover, content filters and PII redaction work out of the box.
- Server-side encryption — As a result, AWS KMS keys protect prompts and outputs.
- CloudTrail logging — Therefore, every InvokeModel call is auditable.
What about the model? GPT-5.5 is OpenAI current frontier model with reasoning, vision, and tool-use. Furthermore, GPT-5.4 ships at the same time as a lower-cost option. Have you ever kept two SDKs in production? In contrast, with Bedrock you can route to either through one API.
Beyond the headline checks, AWS wired in observability hooks. As a result, CloudWatch, X-Ray, and Cost Explorer all work for GPT-5.5 calls.
GPT-5.5 AWS Bedrock Pricing and Availability
So the launch is in limited preview. Therefore, not every AWS account can get access yet. Furthermore, pricing through Bedrock may differ from OpenAI direct rates.
The bigger pricing story is on financial governance. As a result, OpenAI tokens through Bedrock count toward AWS EDPs and Savings Plans. So if your team has annual AWS commits, GPT-5.5 calls apply against that spend.
According to The Register, the move also gives UK and EU public-sector buyers a procurement option. Furthermore, early reports suggest Bedrock rates will track OpenAI direct rates within a few percent.
How to Start Using GPT-5.5 on AWS Bedrock
So getting access follows the standard Bedrock model-access flow with one extra step. Here is the sequence:
- First, open the Bedrock console in a region with the OpenAI models.
- Next, request preview access under Model Access.
- Then, approve the AWS Marketplace agreement.
- After that, wait for the activation email.
- Finally, call the model using the InvokeModel action with the model ID.
- Also, validate logging in CloudTrail.
- Last, wire up cost alarms in CloudWatch.
If you already use Anthropic Claude or Meta Llama on Bedrock, your existing client code works with a single model-ID change. For Bedrock fundamentals, see our AWS Cloud Practitioner certification guide.
GPT-5.5 AWS Bedrock Use Cases
So GPT-5.5 AWS Bedrock unlocks workloads that were awkward on either platform alone. Here are the strongest:
- RAG on private AWS data — Furthermore, pair GPT-5.5 with Kendra or OpenSearch without moving data.
- Agentic workflows — In addition, mix Codex on Bedrock with Managed Agents.
- Multi-model pipelines — Moreover, route easy queries to GPT-5.4 and complex ones to GPT-5.5.
- Regulated industries — As a result, you meet HIPAA, FedRAMP, and SOC 2 with existing AWS attestations.
- Cost-controlled experiments — Therefore, capture token spend in CloudWatch and Cost Explorer.
- Internal dev assistants — So you can point a GPT-5.5 AWS Bedrock pipeline at your codebase.
For event-driven glue, see our AWS Lambda serverless guide.
GPT-5.5 vs OpenAI Direct API
So the choice between GPT-5.5 on Bedrock and the OpenAI direct API comes down to where your data lives. If your stack is on AWS, Bedrock removes integration tax. However, if you are starting fresh, direct API has a slight edge on the latest features.
Bedrock advantages:
- Single SDK for OpenAI, Anthropic, Meta, Mistral, Cohere, and Amazon.
- IAM, PrivateLink, KMS, and CloudTrail apply by default.
- Token spend counts toward AWS commitments.
Direct API advantages:
- New OpenAI features land there first. So Bedrock follows.
- Simpler auth flow if you are not on AWS.
- Direct support and policy escalations.
What would you do? In contrast, most large teams I have worked with run both for the first six months. Then they consolidate.
Summary
So the launch ends Microsoft seven-year OpenAI exclusivity. Therefore, AWS-first teams now get native access to OpenAI frontier models. Furthermore, the integration ships with full IAM, PrivateLink, KMS, and CloudTrail. Access is in limited preview at launch.
Frequently Asked Questions
Is GPT-5.5 AWS Bedrock generally available?
No. So as of the April 28, 2026 announcement, it is in limited preview. Furthermore, GPT-5.4 is available now and GPT-5.5 rolls out within two weeks of the announcement.
Does GPT-5.5 AWS Bedrock cost the same as the direct OpenAI API?
Pricing through Bedrock can differ from OpenAI direct rates. However, token usage counts toward AWS EDPs and Savings Plans. As a result, effective cost can drop for teams with annual AWS commits.
Which AWS regions support GPT-5.5 AWS Bedrock?
At launch, the OpenAI models on Amazon Bedrock live in US East and US West. Furthermore, AWS has not announced a roadmap for more regions.
Can I use Bedrock Guardrails with GPT-5.5?
Yes. So GPT-5.5 AWS Bedrock inherits the standard Bedrock platform features. As a result, content filtering, denied topics, and PII redaction all work.
Does this mean OpenAI is leaving Microsoft Azure?
No. OpenAI models remain on Azure OpenAI Service. In contrast, the April 2026 launch is an expansion, not a replacement.
Editorial Disclosure: This article was researched and drafted with AI assistance. Then it was reviewed, fact-checked, and edited by Bhanu Prakash.
About the Author
Bhanu Prakash is a cybersecurity and cloud computing professional. Furthermore, he has hands-on experience in AWS architecture and Bedrock. He shares practical guides at ElevateWithB.
What to Read Next: Pair this with our AWS Cloud Practitioner certification guide.
