OpenAI brings its frontier models and Codex workflows to Google Cloud

OpenAI and Google Cloud are integrating OpenAI’s frontier models and its Codex coding workflows into Gemini’s Model Garden, according to an announcement published by OpenAI. The models and workflows will be sold through Google Cloud Marketplace, Google’s storefront for third-party software that runs on its cloud.

Model Garden is Google Cloud’s catalog of AI models, a hub where customers can browse, test, and deploy models from multiple developers inside Google’s own cloud infrastructure. Until now, that catalog has centered on Google’s own Gemini models and a mix of open-source and partner models. Adding OpenAI’s frontier models, OpenAI’s term for its most advanced systems, puts a direct rival’s technology inside Google’s own storefront.

Codex is OpenAI’s coding-focused product line. OpenAI’s announcement says Codex workflows are part of what’s moving to Google Cloud, which suggests Google Cloud customers will be able to call on OpenAI’s coding tools without leaving Google’s platform. OpenAI’s announcement does not spell out how these workflows will be packaged or priced once they appear in the Marketplace.

Why a cloud deal between rivals

Google and OpenAI compete directly in AI models, with Gemini and GPT-class systems aimed at many of the same customers. Google Cloud, however, is also one of the infrastructure providers that AI companies depend on to reach enterprise buyers, alongside Microsoft Azure and Amazon Web Services. Making OpenAI’s models available inside Google Cloud Marketplace lets Google sell access to a competitor’s product to its own cloud customers, while giving OpenAI a distribution channel it would otherwise have to build itself. Both companies gain: Google keeps more AI spending inside its cloud billing relationships, and OpenAI reaches customers who have already standardized on Google’s infrastructure.

The arrangement fits a broader pattern in cloud computing, where providers have increasingly hosted models from outside developers alongside their own, competing on distribution and infrastructure rather than insisting customers choose a single model family.

OpenAI’s announcement is short on specifics. It does not name which frontier models are included, when they will go live in Model Garden, what pricing customers will see in the Marketplace listing, or whether the arrangement is exclusive in any way. Google Cloud has not published its own separate statement alongside OpenAI’s, at least none included in the material reviewed here.

What to watch next is whether Google Cloud publishes matching documentation or pricing pages for the Model Garden listing, and whether OpenAI names specific model versions and a rollout date for both the models and the Codex workflows once the integration goes live.