WayToClawEarn
High impactAmazon 官方公告

OpenAI model lands on Amazon Bedrock: Sam Altman and AWS CEO explain the inside story of the cooperation

AWS and OpenAI announced the expansion of their cooperation: OpenAI’s cutting-edge models will be available on Amazon Bedrock for the first time, Codex programming agents will be launched simultaneously, and Bedrock Managed Agents, powered by OpenAI, will be launched. Enterprises can use models such as OpenAI and Anthropic simultaneously on a single infrastructure.

WayToClawEarn EditorialPublished Apr 29, 2026Updated Aug 8, 2026

Editorial review of public sources · AI-assisted drafting. How we work · Original source

Core conclusion

On April 28, 2026, AWS and OpenAI announced a major expansion of cooperation: OpenAI cutting-edge models (including the GPT-5 series) will be available through Amazon Bedrock for the first time, and the Codex programming agent will also be available on Bedrock. More importantly, the two parties jointly launched Amazon Bedrock Managed Agents, powered by OpenAI - a managed Agent service that runs OpenAI models on AWS infrastructure. What this means for enterprise AI application teams is that they no longer have to choose between "the best model" and "the most reliable cloud infrastructure."

Key Points

  • Time of incident: 2026-04-28 -Affected objects: Enterprise AI teams, AI Agent developers, and content automation pipeline operators using AWS
  • Core changes: OpenAI models are officially available on AWS Bedrock. Enterprises can use OpenAI, Anthropic Claude, Meta Llama and other models simultaneously in one service

Background and trigger events

Ben Thompson of Stratechery exclusively interviewed OpenAI CEO Sam Altman and AWS CEO Matt Garman, and simultaneously released three major cooperation contents:

  1. OpenAI models on Bedrock: The latest OpenAI models such as the GPT-5 series can be called through the Bedrock API, and unified use of enterprise-level security controls such as AWS IAM permission management, PrivateLink network isolation, and CloudTrail audit logs
  2. Codex Programming Agent on Bedrock: 4 million+ users use Codex-assisted programming every day, and now enterprise teams can directly access Codex CLI, desktop and VS Code extensions in the AWS environment
  3. Bedrock Managed Agents, powered by OpenAI: Managed Agent service that integrates the inference capabilities of OpenAI’s cutting-edge models with the global infrastructure of AWS

AWS’ official statement emphasizes that the OpenAI model inherits a complete enterprise control system on Bedrock: IAM access management, AWS PrivateLink connections, encryption (in transit and at rest), and AWS CloudTrail comprehensive logging.

Key Impact (by Dimension)

DimensionsChangesWhat it means for enterprise AI teamsRecommended actions
CostOpenAI usage can be counted toward existing AWS cloud commitmentsSimplified procurement process, unified management of AI spending and AWS billingEvaluate whether existing AWS Reserved Instances cover new OpenAI calls
SecurityUnified IAM permissions + PrivateLink + encryption at rest/in-transitNo more separate secure channels for OpenAI, audit integrationManage OpenAI models into existing Bedrock Guardrails policies
Model selectionBedrock also provides OpenAI + Anthropic + Meta + Cohere + AmazonTruly selecting the optimal model for each type of task without sacrificing infrastructure consistencyComparing the performance differences between OpenAI and Claude in RAG scenarios
Agent capabilitiesManaged Agents come with persistent memory, skill encoding, and identity permissionsThe work of building Agent infrastructure from scratch is greatly reducedMigration assessment of existing Agent processes

Adaptation suggestions

For readers who are concerned about making money through AI automation, this cooperation brings several direct implementation directions:

Cloud infrastructure upgrade for AI content workflows

  • If you already have an automated content pipeline running on AWS, you can now call OpenAI models directly in the Bedrock API without additional configuration of API Keys, network policies, and billing channels.
  • Migration plan: Switch the direct call to the OpenAI API in the Python script to the Bedrock API call (using boto3 SDK), inheriting the current limiting, retry and logging capabilities of AWS

Use Codex to improve AI Agent code production

  • Codex is used by 4 million+ developers every week for code auto-completion, refactoring and test generation
  • Combined with Bedrock Managed Agents, you can build a complete closed loop in which AI Agent calls Codex to write code, automatically deploys to AWS Lambda, and monitors operation.

Automated Applications of Bedrock Managed Agents

  • Managed Agents come with persistent memory (maintaining context across sessions) and skill encoding (reusable workflows)
  • Suitable for building AI workflows that need to remember which step you did last time, such as a complete Pipeline for automatic content collection, processing and publishing

Example: Call OpenAI model using Bedrock API

python
import boto3
import json

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

response = bedrock.invoke_model(
    modelId='openai.gpt-5',
    contentType='application/json',
    accept='application/json',
    body=json.dumps({
        "messages": [
 {"role": "system", "content": " SEO AI 。"},
 {"role": "user", "content": " 100 OpenAI AWS 。"}
        ],
        "max_tokens": 200,
        "temperature": 0.7
    })
)

result = json.loads(response['body'].read())
print(result['choices'][0]['message']['content'])

Bedrock API code example

AI

OpenAI Bedrock,。 Bedrock Managed Agents, powered by OpenAI

AWS Managed Agents (persistent memory)、(skills)(identity)。 2026 AI Agent 。

AI ,

  • Agent 、,
  • OpenAI Agent, AWS
  • Anthropic Claude Bedrock , OpenAI Claude Agent

OpenAIChatGPTClaudeClaude Coden8nAWSDeepSeekHermes Agent.

Internal link guidance

View source →

Disclaimer: this site shares educational insights only, for inspiration and reference. No outcome guarantee; external execution and decisions are your own responsibility.