OpenAI Puts Frontier Models on Google Cloud's Model Garden

Summary

• OpenAI said its frontier models and Codex agentic workloads are coming to Google Cloud's Model Garden.
• Transactions run through Google Cloud Marketplace, with purchases payable from existing cloud spend commitments.
• The deal widens OpenAI's enterprise distribution even as Microsoft remains its primary cloud partner.

The latest

OpenAI's frontier models and Codex agentic workloads will be sold through Google Cloud's Model Garden, with transactions handled on Google Cloud Marketplace, the company announced on October 21, 2026. OpenAI said Google Cloud customers gain native access to the models inside agentic workloads, and that general availability is planned in the coming weeks.

Details

• The mechanics: According to OpenAI, the models and Codex agentic workloads are listed in Model Garden, Google Cloud's catalog of available models, while the commercial transaction itself is completed through Google Cloud Marketplace. OpenAI described the result as native access to its models within customers' agentic workloads.
• The commercial hook: OpenAI said enterprises holding existing spend commitments with Google Cloud will be able to apply those commitments to purchases of its models. That removes a separate procurement cycle for companies that have already contracted cloud budget, and places OpenAI inside an established enterprise buying channel.
• The governance pitch: The integration pairs the models with what OpenAI described as built-in access controls, centralized data governance and high-performance infrastructure. The stated aim is to help organizations in regulated sectors control who reaches sensitive data and how that data is used.
• Timing: OpenAI said general availability is planned for the coming weeks. Access in the meantime runs through a request form on the announcement page, meaning entry is gated rather than open to every Google Cloud account from day one.
• What wasn't specified: OpenAI did not name specific model versions covered by the arrangement, referring only to its frontier models in general terms. No pricing, no customer names and no fixed launch date were attached to the announcement.
• Wider rollout: OpenAI added that its models will also become available soon inside Google Antigravity and Gemini Agent, extending the tie-up beyond the Model Garden listing into Google's own agent-facing products.
• The Microsoft line: OpenAI stated that Microsoft remains its primary cloud partner. The explicit reassurance signals the Google arrangement is a distribution channel rather than a change in where OpenAI's core infrastructure sits.
• Who it targets: The offer is aimed at regulated-sector enterprises, where data handling rules typically slow adoption of external models. OpenAI framed centralized governance and access controls as the feature set addressing that barrier.

Background

Model Garden is Google Cloud's catalog for discovering and deploying models, and Marketplace is its billing channel, where committed enterprise spend can be drawn down. Listing a rival lab's frontier models there puts OpenAI in front of Google's enterprise base.

Between the lines

The commercial design matters more than the listing. Letting customers pay with existing Google Cloud commitments converts budget already locked to one vendor into OpenAI revenue, which is why the Microsoft reassurance sits in the same announcement. The absence of named model versions and pricing, alongside a request-form gate, suggests a staged enterprise rollout rather than a full open launch.

What's next

Watch for general availability in the coming weeks, the specific model versions OpenAI ultimately lists, pricing terms on Google Cloud Marketplace, and the timing of availability inside Google Antigravity and Gemini Agent.

Source: OpenAI