OpenAI frontier models hit Bedrock GA, putting OpenAI and Anthropic side by side on AWS
OpenAI's GPT-5.5, GPT-5.4, and Codex reached general availability on Amazon Bedrock on 2026-06-01, six weeks after the Microsoft cloud exclusivity ended. AWS is now the only platform where the OpenAI and Anthropic frontiers sit behind the same console, IAM, and billing.
OpenAI moved its GPT-5.5, GPT-5.4, and Codex coding agent to general availability on Amazon Bedrock on 2026-06-01, one month after a limited preview and six weeks after OpenAI’s exclusive cloud distribution agreement with Microsoft formally ended on 2026-04-27. The headline is straightforward. The procurement geometry the launch creates is the part that reshapes how AI buying decisions get made in the next quarter.
What shipped
GPT-5.5 lands in the US East (Ohio) Bedrock region. GPT-5.4 is available in US East (Ohio) and US West (Oregon), with Commercial and GovCloud coverage on the roadmap. Pricing matches OpenAI’s first-party rates with no Bedrock markup: GPT-5.5 at $5 per million input tokens and $30 per million output, GPT-5.4 at $2.50 and $15. Codex on Bedrock ships through the Codex App, the Codex CLI, and the VS Code, JetBrains, and Xcode integrations, with all inference routed inside the customer’s chosen region for data residency. Billing rolls into existing AWS commitments, and calls inherit the standard Bedrock controls: IAM permissions, VPC and PrivateLink isolation, KMS encryption, CloudTrail audit logs, and a contractual guarantee that prompts and responses are not used to train models.
AWS also previewed Amazon Bedrock Managed Agents powered by the OpenAI agent harness, slated for general availability later, and confirmed that OpenAI’s Daybreak cybersecurity initiative is the next OpenAI capability headed to Bedrock.
Where this lands in the market
The substance is the procurement angle, not the inference. AWS Bedrock is now the only major cloud that hosts both OpenAI’s GPT-5.5 line and Anthropic’s Claude Opus 4.8 line under the same console, the same IAM model, and the same per-token billing. Azure has Anthropic in Foundry but the OpenAI integration is older and deeper; Google Cloud ships Anthropic on Vertex AI but does not host OpenAI. CIOs who have been running side-by-side evaluations on Claude and GPT through two separate vendor contracts can collapse that to one procurement relationship with one set of compliance documentation.
That changes the rate at which teams can re-evaluate the frontier. Switching from GPT-5.5 to Claude Opus 4.8 for a given workload used to involve a fresh SOC 2 review and a new data-processing addendum. Inside Bedrock, it becomes a model ID change with the same audit trail. The companies that win in this configuration are the ones whose buyers already prefer one cloud governance surface over many; AWS-native shops were already overweight Anthropic and will now find the OpenAI path materially cheaper to adopt than going through Microsoft.
The pricing parity is also a pricing signal. OpenAI’s per-token rates on Bedrock match its direct API to the cent, which means OpenAI is not discounting cloud-mediated distribution to drive volume. That is a different posture than the early Bedrock launches from Mistral and Cohere, where the cloud relationship came with margin compression. The read is that OpenAI is treating AWS as a complementary distribution channel, not a wholesale partner, and is comfortable carrying its first-party pricing into a competitor cloud.
What’s worth watching
-
Bedrock Managed Agents launch terms. AWS previewed an agent runtime “powered by OpenAI” but did not commit to GA timing or pricing. If Managed Agents undercut the equivalent OpenAI Assistants pricing, it will be the first sign that Bedrock margin pressure reaches the agent layer, not just the model layer.
-
Anthropic’s response on Azure. Anthropic shipped Opus 4.8 into Azure AI Foundry in May. If Anthropic now broadens the available regions or surfaces Claude Code inside Foundry to mirror Codex on Bedrock, it confirms that both labs are treating cloud-mediated distribution as table stakes rather than an experiment.
-
Daybreak’s positioning. OpenAI named Daybreak as the next capability landing on AWS. Daybreak is a cybersecurity initiative, not a model: that is a clue OpenAI is willing to ship vertical workflows through Bedrock, not only inference. If it lands with an enterprise security marketing motion, expect the competitive overlap with Anthropic’s Enterprise Agents and AWS Security Hub to get loud quickly.
The next checkpoint is the Bedrock Managed Agents GA. The cloud distribution story for the OpenAI frontier line is no longer Azure-only, and the second-order question is how aggressively AWS uses that to widen the agent-layer surface beneath the model layer.
- AWS: Get started with OpenAI GPT-5.5, GPT-5.4 models, and Codex on Amazon Bedrock aws.amazon.com
- OpenAI: OpenAI frontier models and Codex are now available on AWS openai.com
- About Amazon: OpenAI GPT-5.5, GPT-5.4, and Codex now generally available on Amazon Bedrock www.aboutamazon.com
- AWS ML Blog: OpenAI models and Codex on Amazon Bedrock are now generally available aws.amazon.com