OpenAI Models Join Google Cloud’s Model Garden for Enterprise Work

OpenAI and Google Cloud announced on 21 October that OpenAI models and Codex agent workflows are coming to Google Cloud’s Model Garden. The partnership gives Google Cloud customers another way to access OpenAI models through infrastructure and procurement they already use.[1][2]

The announcement is at preview stage. Google says access is rolling out to selected enterprise customers now; Google gives no general-availability date. Its post also lists no final prices and no Malaysia-specific access terms.[1][2]

Google Cloud says its Model Garden will include GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. Customers will be able to call the models through Google Cloud APIs, while Codex CLI and desktop workflows can connect to Google’s Gemini platform for inference. Google also says future access through Google Antigravity and Gemini for business is planned.[2]

The commercial route may matter as much as the model list. Google says purchases will be transacted through Google Cloud Marketplace, appear on the customer’s Google Cloud invoice and count against eligible cloud spending commitments. Google adds that usage also counts toward Flexible Savings Plans across model providers.[2][1]

Google describes the current phase as a private preview for selected enterprise customers, with more access to follow. Its post says customers can request access and begin building through APIs. Google gives no general-availability date, so smaller firms should not assume that they can sign up today or that every account is eligible.[2][1]

The companies have not published final per-token rates, usage limits, support terms or a fixed general-availability date in these announcements. A Google Cloud Marketplace transaction can simplify billing for an existing customer, but it does not by itself show that the combined service will cost less than buying model access another way.[2][1]

For a small team, one invoice is a procurement feature, not a performance or savings result. Before a switch, compare response time, model quality, cloud transfer charges, request limits and support with the current route. These checks matter most when a workload includes customer data or supports a time-critical process.[2][1]

Google says prompts, context and outputs will stay within the customer’s designated data perimeter and will not be used to train foundation models. It also describes authenticated identities, audit logging and network controls. These are Google’s statements about the service; procurement teams should compare them with the binding terms, model-specific documentation and settings available in their own project.[2]

Zero Data Retention and Confidential Compute are described as coming soon, not as features available in this preview. The Google arrangement adds a route for Google Cloud customers, but the companies have not said that every model, feature or data-control option will be available on the same timetable.[2][1]