OpenAI GPT-5.5 Is Now on Amazon Bedrock

AWS and OpenAI just partnered. GPT-5.5 and Codex are coming to Bedrock — same APIs, same security, same billing. Here's what changes.

meghana rajagopal

Published Apr 30, 2026

Last modified Apr 30, 2026

OpenAI Models Are Now on Amazon Bedrock — What This Changes for Your AI Stack

TL;DR: AWS and OpenAI just announced a landmark partnership. GPT-5.5 and Codex are coming to Amazon Bedrock — same APIs, same security, same billing. Here's what it means, what's available now, and how to start using it today.

The Announcement Nobody Saw Coming

On April 28, 2026, AWS and OpenAI confirmed a $50 billion partnership that ends OpenAI's exclusivity with Microsoft Azure.

The headline: OpenAI's frontier models — including GPT-5.5 — are coming to Amazon Bedrock.

For anyone building AI on AWS, this is a fundamental shift. You no longer have to choose between AWS's infrastructure and OpenAI's models. You get both — through the same APIs you already use.

What's Actually Available

1. OpenAI Models on Bedrock (Limited Preview)

GPT-5.4 is available now in preview. GPT-5.5 is coming within weeks. Access them through the standard Bedrock InvokeModel API — no new infrastructure, no new security model, no separate OpenAI account needed.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
import boto3

import json



bedrock = boto3.client('bedrock-runtime', region_name='us-east-1')



# Call GPT-5.4 via Amazon Bedrock — same API as Claude or Titan

response = bedrock.invoke_model(

    modelId='openai.gpt-5-4',  # OpenAI model via Bedrock

    body=json.dumps({

        "messages": [

            {

                "role": "user",

                "content": "Explain the difference between RAG and fine-tuning in one paragraph."

            }

        ],

        "max_tokens": 500,

        "temperature": 0.7

    }),

    contentType='application/json',

    accept='application/json'

)



result = json.loads(response['body'].read())

print(result['choices'][0]['message']['content'])

What stays the same:

AWS credentials (IAM roles, no separate OpenAI API key needed)
Bedrock's cost controls and budget alerts
CloudWatch logging and monitoring
VPC endpoints and data residency controls
Billing through your existing AWS account

2. Codex on Amazon Bedrock

OpenAI's Codex — the coding-focused model — is now available within your AWS environment. This means:

Authenticate with AWS credentials
Inference runs through Bedrock infrastructure
Usage counts toward your AWS cloud commitments

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
import boto3

import json



bedrock = boto3.client('bedrock-runtime', region_name='us-east-1')



# Use Codex for code generation via Bedrock

response = bedrock.invoke_model(

    modelId='openai.codex-1',

    body=json.dumps({

        "messages": [

            {

                "role": "system",

                "content": "You are an expert AWS Python developer."

            },

            {

                "role": "user",

                "content": "Write a Lambda function that reads from S3, processes records, and writes results to DynamoDB."

            }

        ],

        "max_tokens": 1000

    }),

    contentType='application/json',

    accept='application/json'

)



result = json.loads(response['body'].read())

print(result['choices'][0]['message']['content'])

3. Amazon Bedrock Managed Agents Powered by OpenAI

This is the most exciting piece for production builders. Bedrock Managed Agents now supports OpenAI frontier models as the reasoning engine — combining:

OpenAI's model quality and agentic capability
AWS's infrastructure, session management, and tool execution
The same Action Groups and Knowledge Bases from your existing Bedrock setup

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
import boto3



bedrock_agent = boto3.client('bedrock-agent', region_name='us-east-1')



# Create a Bedrock Agent powered by GPT-5.4

agent = bedrock_agent.create_agent(

    agentName='openai-powered-support-agent',

    agentResourceRoleArn='arn:aws:iam::123456789012:role/AmazonBedrockAgentRole',

    foundationModel='openai.gpt-5-4',  # OpenAI model as the brain

    instruction="""You are a helpful customer support agent.

    Look up orders, create tickets, and send confirmations.

    Be concise and professional."""

)



agent_id = agent['agent']['agentId']

print(f"Agent created with GPT-5.4: {agent_id}")

Your existing Lambda Action Groups, OpenAPI schemas, and Knowledge Base attachments all work — just swap the model.

Claude vs GPT-5.4 on Bedrock — Which to Use?

Now that both are available through the same API, here's a practical guide:

Use Case	Recommended Model
Document Q&A / RAG	Claude 3 Haiku (cheaper, fast)
Complex reasoning / analysis	Claude 3 Sonnet or GPT-5.4
Code generation	Codex or Claude Sonnet
Agentic tasks (multi-step)	GPT-5.5 or Claude 3 Sonnet
Cost-sensitive high volume	Claude 3 Haiku
Existing OpenAI workflows migrating to AWS	GPT-5.4 (no code changes needed)

The practical answer: start with Claude for new builds (better Bedrock integration, mature tooling). Use GPT-5.4/5.5 when migrating existing OpenAI workloads to AWS — it's a drop-in path with no model changes required.

Why This Matters for Enterprise AWS Teams

1. No More Cloud Switching for AI

Teams who built on OpenAI and wanted AWS infrastructure had to manage two clouds, two billing systems, two security models. That's gone.

2. Unified Governance

Bedrock's Guardrails, CloudWatch logging, and VPC endpoints now apply to OpenAI models too. One compliance framework covers your entire AI stack.

3. Negotiating Power

Your AWS Enterprise Discount Programme (EDP) now covers OpenAI model usage. OpenAI consumption counts toward AWS spend commitments — significant for large enterprises.

How to Get Access

The partnership is currently in limited preview. To request access:

Go to Amazon Bedrock → Model access in your AWS console
Look for OpenAI models in the model catalogue
Request access — approval is typically within a few days for AWS Partner organisations

1
2
3
4
5
# Check available models including OpenAI via CLI

aws bedrock list-foundation-models \

  --region us-east-1 \

  --query 'modelSummaries[?contains(modelId, `openai`)].[modelId,modelName]' \

  --output table

My Take

This partnership doesn't replace Claude on Bedrock. It expands your options.

For net-new AWS AI projects, I still recommend starting with Claude — the Bedrock Knowledge Bases integration, Guardrails support, and pricing at scale are hard to beat.

But for enterprise teams with existing OpenAI investments who want to consolidate onto AWS infrastructure — this is the migration path they've been waiting for.

The real winner here is the customer. More model choice, one infrastructure, unified governance.

Resources

'm Meghana Rajagopal, lead Cloud & AI Delivery at STC (AWS Premier Partner) and AWS ML Engineer Associate.This is Part 5 of my series on building production AI on AWS. I write about practical AWS AI/ML solutions and architecture. Connect with me on LinkedIn .

# amazon-bedrock # openai-gpt # generative-ai

Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.

Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article

Builder Center

OpenAI GPT-5.5 Is Now on Amazon Bedrock

meghana rajagopal

OpenAI Models Are Now on Amazon Bedrock — What This Changes for Your AI Stack

The Announcement Nobody Saw Coming

What's Actually Available

1. OpenAI Models on Bedrock (Limited Preview)

2. Codex on Amazon Bedrock

3. Amazon Bedrock Managed Agents Powered by OpenAI

Claude vs GPT-5.4 on Bedrock — Which to Use?

Why This Matters for Enterprise AWS Teams

1. No More Cloud Switching for AI

2. Unified Governance

3. Negotiating Power

How to Get Access

My Take

Resources