OpenAI models and Codex workflows are coming to Google Cloud's Model Garden

OpenAI says its frontier models and its Codex agentic workflows are coming to Google Cloud's Model Garden. Google Cloud customers will be able to buy them through the Google Cloud Marketplace. Companies that have spend commitments with Google Cloud will be able to count these purchases against those commitments. OpenAI says general availability is planned "in the coming weeks."

The news is less about what the models can do and more about where companies buy and run them. A company that already keeps its data, permissions and budget in Google Cloud will be able to add OpenAI models without leaving that environment. Google builds its own Gemini models, so it will now also sell a direct competitor's models next to them.

What was announced

According to OpenAI's announcement, the partnership has three parts.

• Frontier models in Model Garden. OpenAI says customers get "native access" to its frontier models inside Google Cloud, to run "complex agentic workflows at scale."
• Codex workflows. OpenAI's agentic Codex workflows come along with the models. OpenAI links this to enterprises that want to "modernize their software development lifecycle" and build agentic workflows.
• Google's own agent products. OpenAI says its models will "soon" be available in Google Antigravity and the Gemini Agent. It presents this as giving enterprise customers model choice and "the best model for each task."

The post leaves several things out. It does not say which OpenAI models will be offered, what they will cost, or how Codex will run inside Google's environment. It gives no figures on expected usage or revenue, and no exact date. For now, customers are asked to request access.

Buying through the cloud bill

The commercial detail matters most here. Large companies often agree to spend a fixed amount with a cloud provider over a set period. OpenAI says enterprises with Google Cloud spend commitments "will be able to draw down their GCP commitments" when they buy OpenAI models through the Marketplace.

In practice, a team that wants OpenAI models no longer needs a separate supplier contract, a new vendor review or a new budget line. The purchase goes through the account the company already has with Google. OpenAI describes this as "a path from evaluation to broader adoption." That is a fair description of how purchasing works in large companies: tools that fit an existing contract get trialled sooner and spread faster.

It also means Google Cloud handles the transaction. OpenAI still makes the model, but the customer relationship, the bill and the procurement route run through Google's marketplace.

Governance and who holds the controls

OpenAI aims the announcement at regulated industries. It says that deploying AI there "requires control over who can access sensitive data and how it is used." The integration will pair OpenAI's models with Google Cloud's "built-in access controls, centralized data governance, and high-performance infrastructure," in OpenAI's words.

This splits control in three. OpenAI supplies the model. Google Cloud supplies the permissions, the data governance rules and the compute the work runs on. The customer chooses, model by model and task by task, which provider to use inside one platform. Bringing OpenAI models into Antigravity and the Gemini Agent goes a step further: Google's own agent tools would let users pick an OpenAI model where they might otherwise have used a Gemini one.

For a company's IT and security teams, the effect is that access policy and data rules can stay in one place even when the models come from several vendors. The model becomes a part that can be swapped. The platform around it stays fixed.

What it means for work

All of this changes things first for the people who choose and approve AI tools: procurement, IT and security teams. They will be able to add OpenAI models under the contract, controls and audit setup they already run in Google Cloud. For software teams, Codex workflows inside Google Cloud could put agentic coding tasks, such as writing and changing code, into the same environment where their code and data already live. Developers using Antigravity or the Gemini Agent may soon pick a model per task instead of being tied to one vendor. Several questions are still open: which models and Codex features will actually ship, at what price, and how Google's governance controls will apply to agents that act on a company's systems. For now, the announcement says more about who sells and controls access to AI tools than about what those tools can do.