GPT-5.5 and GPT-5.4 now available in Amazon Bedrock

Amazon has made OpenAI's GPT-5.5, GPT-5.4 and Codex models generally available in its Amazon Bedrock cloud service, the AWS Machine Learning Blog reported. Developers can already deploy these models in production applications and agent systems on top of Bedrock's high-performance inference engine, without leaving AWS infrastructure.

What is now available in Bedrock

Amazon Bedrock is a managed AWS service that gives developers a single API for working with models from different providers, from Amazon's own models to external models from other major AI labs. The addition of OpenAI's models to this list, GPT-5.5, GPT-5.4 and Codex, a model specialized in writing code, expands Bedrock's already broad multi-model catalog. Companies that already build their infrastructure on AWS can use OpenAI's top models without integrating separately with the OpenAI API or maintaining separate billing outside the AWS perimeter.

Key facts about the launch:

• Models: GPT-5.5, GPT-5.4 and Codex
• Status: generally available in production
• Platform: Amazon Bedrock, an AWS cloud service
• Purpose: deployment in production applications and agent systems
• Infrastructure: runs on top of Bedrock's high-performance inference engine

Why this matters for the cloud AI market

Until recently, access to OpenAI's flagship models almost always meant working directly with the OpenAI API or using Microsoft Azure, OpenAI's long-time technology partner. The arrival of GPT-5.5, GPT-5.4 and Codex in Bedrock means AWS is cementing its multi-model cloud strategy: rather than betting on a single model provider, Amazon gives customers a choice between different model families within one infrastructure, keeping the role of a neutral platform rather than an exclusive access channel to one ecosystem.

For enterprise customers, this lowers the barrier to entry: there is no need to go through a separate security and compliance certification for a third-party OpenAI API if the company's infrastructure is already built on AWS and has passed all the checks required to work with Bedrock. This matters especially in regulated industries such as finance, healthcare and government, where any new external connection to company data requires a separate security audit, and a model appearing inside an already approved cloud perimeter removes a significant part of the bureaucratic burden. AWS has applied the same logic before to other external models in the Bedrock catalog: the company consistently adds top models from different labs as they are released, reinforcing its reputation as the cloud where enterprise customers get access to the market's most advanced work without having to negotiate and sign contracts with each model provider individually.

What this changes for developers

The practical effect for teams building AI agents and products on large language models is the ability to compare and switch between models from different providers through Bedrock's single interface, without rewriting the application's integration layer. Codex is of particular significance: it is a model optimized for writing and refactoring code, and its arrival in Bedrock opens the way to embedding agentic development tools directly into enterprise CI/CD pipelines on AWS, without moving code outside the customer's cloud.

Combined with growing competition between Bedrock and other major cloud platforms, this reinforces the overall trend of 2026: hyperscalers are competing less over exclusivity of their own models than over becoming the universal access point to all of the market's leading models at once, including models from direct competitors in the infrastructure business.

The symbolic weight of the move also deserves a mention: Amazon, which develops its own Nova model family and has also invested in Anthropic, a competitor of OpenAI, nevertheless hosts models from OpenAI, Anthropic's biggest rival in the frontier-model market. This highlights that for Amazon's cloud business, the commercial logic of multi-vendor support outweighs any strategic alliance around a particular AI lab: for a Bedrock customer, having a choice among the best available models here and now matters more than seeing only models from companies friendly to the platform's owner.