News

Releases & Models

CONFIRMED: OpenAI brings Codex and its GPT-6 models to Google Cloud, the third hyperscaler to sell the brand

OpenAI and Google Cloud announced a partnership to bring GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna, plus Codex workflows, into Google's Model Garden. General availability lands in the coming weeks, but preview access already started.

Ilustração editorial de torres de servidores de data center interligadas dentro de uma nuvem azul luminosa, representando os modelos da OpenAI entrando no Model Garden do Google Cloud

OpenAI stops betting only on Microsoft

OpenAI confirmed, in an official announcement published this Wednesday (October 21), that its frontier models and Codex, the company's coding agent, will start running inside Google Cloud. The integration happens through Model Garden, the showcase where Google Cloud already offers hundreds of AI models to its enterprise customers, including Gemini itself. With this, Google becomes the third major cloud, after Microsoft Azure and Amazon Web Services, to commercially sell OpenAI's models to businesses.

What changes in practice

The GPT-6 Astra model (the most powerful in the family, aimed at complex coding, computer use and scientific research tasks), GPT-6.1 Sol (a cheaper version nearly as capable as Astra) and GPT-6 Luna (the fastest and most cost efficient, built for everyday work at scale) enter private preview for select customers. OpenAI says general availability arrives in the coming weeks. Beyond the models, anyone already using Codex locally, whether through the command line or the desktop app, will be able to connect those workflows directly to Google's infrastructure for inference.

Why this matters to cloud buyers

The most practical part of the announcement is commercial, not technical. Google Cloud customers will be able to purchase OpenAI's models through the Marketplace, with consolidated billing on a single Google invoice, and can draw down existing Google Cloud spend commitments to pay for that usage. Consumption also counts toward Flexible Savings Plans that span different model providers. In practice, a marketing or automation company that already has budget committed to Google Cloud gains access to OpenAI's models without opening a new account, signing a separate contract, or managing two distinct bills. That lowers adoption friction for teams already living inside the Google ecosystem who want to test or run OpenAI models in production without extra bureaucracy.

Governance was Google's central argument

Google Cloud published its own announcement the same day, emphasizing that OpenAI's models will run inside isolated data perimeters, without using customer prompts and outputs to train anyone's models, with dedicated infrastructure and unified identity auditing. Features like zero data retention and confidential computing appear as 'coming soon,' meaning they are not yet available at launch. This is Google's bet to attract companies in regulated sectors, such as healthcare and financial services, that need strict control over where data travels before authorizing any third-party AI use.

The context: Microsoft stopped being exclusive

This move is only possible because OpenAI renegotiated its exclusivity agreement with Microsoft Azure earlier in 2026. Since then, the company had already expanded to Amazon Web Services, first offering its open weight models and later its frontier models and Codex itself in limited preview. Google Cloud is repeating, for OpenAI, the move it had already made shortly before with Anthropic, assembling a catalog that today brings together the three most relevant closed source AI labs inside a single platform. For Google, this reinforces the narrative that Model Garden is neutral: customers choose the model for the task, not for the cloud provider.

The limit that still exists

It's worth separating the announcement from actual delivery. Today access is private preview, restricted to select customers, with no exact general availability date beyond 'in the coming weeks.' More sensitive security features, like zero data retention and confidential computing, are not live yet. And OpenAI itself makes a point of saying Microsoft remains its primary cloud partner, signaling that this Google deal expands distribution without breaking the deeper tie the company keeps with Azure.

MaxAssistant's take

The real message here isn't about a new model, it's about distribution. OpenAI is moving away from depending on a single cloud to reach enterprise customers and shifting toward a logic of being available wherever the buyer already spends money, whether Azure, AWS or now Google Cloud. For mid sized companies doing automation, marketing or customer service that already use Google Cloud for other things, this deal removes a real adoption barrier: fewer contracts, fewer vendors, more model choice within the same invoice. That tends to accelerate production testing with GPT-6 among teams that previously avoided opening a new account just to experiment.

Sources

OpenAI: OpenAI models are coming to Google Cloud; Google Cloud Blog: Google Cloud expands model choice and flexibility, bringing OpenAI models to customers; CNBC: OpenAI brings its AI models to Google's cloud, expands beyond Microsoft.