OpenAI officially launches on AWS Bedrock: Codex, cutting-edge models, and managed agents are fully open to enterprise-level access

Core conclusion

On April 28, 2026, AWS and OpenAI announced the largest cooperation upgrade in history: OpenAI cutting-edge models (including open weight inference models such as gpt-oss-120b and gpt-oss-20b) are officially available on Amazon Bedrock. This is not just a model launch, but AWS has created a complete "enterprise-level base" for OpenAI - Codex coding agents and Bedrock Managed Agents are simultaneously open in Limited Preview form.

For teams using AI for automation, content production, and tool development, this means three important changes:

1. Farewell to the Azure Exclusive Era: OpenAI models are now available on AWS, managed under the same API as Anthropic Claude, Meta Llama and other models
2. Codex comes to the enterprise: The coding agent used by 4 million people every week is now available through AWS certification and billing system
3. Managed Agents lowers the threshold: AWS helps you manage memories, skills, and permissions on Bedrock - you only need to define what the Agent does

Key Points

• Time of incident: 2026-04-28 -Affected objects: All teams that use AI models for automation and content production
• Core changes: OpenAI fully opens AWS Bedrock enterprise-level access

Background and trigger events

OpenAI's models have previously mainly used its own API and Azure exclusive channels. And Anthropic quickly penetrated a large number of enterprise customers in 2025 through Bedrock. HN community comments directly stated:

"Anthropic became the workhorse model for my organization through Bedrock...OpenAI provides such a poor enterprise experience on Azure that they are almost excluded from enterprise-grade deployments."

This collaboration is a direct response to this situation. AWS and OpenAI CEOs jointly interviewed Stratechery to announce three simultaneous releases:

1. OpenAI Models on Amazon Bedrock — Latest OpenAI models served via the Bedrock API
2. Codex on Amazon Bedrock — OpenAI Coding Agent is available for enterprise-level software development
3. Amazon Bedrock Managed Agents, powered by OpenAI — Exclusive managed agent experience

Key Impact (by Dimension)

Adaptation suggestions

1. Evaluate your current AI tool stack

If your team uses both OpenAI APIs and AWS services, now is the perfect time to move to Bedrock Unified Gateway. IAM permission management, VPC private connections, and CloudTrail log auditing can cover all models in one stop.

2. Test Codex on Bedrock

Codex, which already has 4 million weekly active users, is now available through the Bedrock API. If the team is using coding agent tools such as Claude Code and Copilot, the performance of Codex can be tested in parallel, especially in complex multi-step reasoning and cross-file reconstruction.

3. Pay attention to the opening rhythm of Managed Agents

Managed Agents are currently in Limited Preview status. If your business needs to build multiple AI Agents (such as content collection agents, data cleaning agents, release review agents), the memory persistence and skill encoding capabilities of Managed Agents will significantly reduce development costs.

4. Multi-model strategy

Bedrock now supports OpenAI, Anthropic, Meta, Mistral, Cohere, Amazon and other models. It is recommended to establish "model routing strategies" for different task types: OpenAI's gpt-oss-120b is suitable for inference tasks, Anthropic is suitable for content production tasks, and small models are suitable for classification and preprocessing.

Task List

• Request OpenAI on Bedrock Limited Preview (via AWS console)
• Add Bedrock as a model gateway option in existing AI workflows
• Test the Codex CLI integration with the AWS SDK to see if it can reduce manual code migrations
• Pay attention to the official release of Managed Agents and evaluate whether to replace the existing Agent framework

Example: Call OpenAI model through Bedrock API

# AWS CLI OpenAI （ Bedrock ）
aws bedrock-runtime invoke-model \
  --model-id "openai.gpt-oss-120b-v1" \
  --body '{"prompt": "Write a Python script to parse CSV files","max_tokens": 500}' \
  --cli-binary-format raw-in-base64-out \
  /tmp/invoke_output.json

#
cat /tmp/invoke_output.json | python3 -m json.tool

• AWS ：OpenAI AWS
• Amazon ：AWS and OpenAI expanded partnership
• Stratechery ：OpenAI AWS CEO

（）

， tools OpenAI、ChatGPT、Claude、Codex、Anthropic、n8n、LangGraph

Internal link guidance

• Want to practice AI automated workflow? See: n8n + OpenAI ：30
• Real case: Indie Developer: n8n + OpenClaw Automation Workflow Earning $5,000/mo
• AI Agent Entrepreneurship: Claude Code 48 hours to start a business: one person + US$29 monthly fee, monthly income in 3 months $9,000