OpenAI's GPT-6 models arrive on Google Cloud

The cloud you already pay for is no longer a reason to skip OpenAI.

Google Cloud says OpenAI's GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are rolling out in private preview to select enterprise customers in its model garden. Usage bills through Google Cloud Marketplace, on the same invoice as everything else you run there.

Three GPT-6 models, one Google invoice

Google Cloud's announcement, posted October 21, lists the three models and what each is for. Astra is pitched for demanding work in coding, computer use and research. Sol offers "near-Astra intelligence at lower cost." Luna is the fast, cheap one for everyday volume.

The billing detail matters more than the model list. Google says OpenAI usage "draws down existing cloud commitments" and counts toward its Flexible Savings Plans across model providers.

On data, Google says prompts, context and outputs stay inside your designated data perimeter and are never used to train foundation models. Zero Data Retention and Confidential Compute are both listed as "coming soon," with no dates.

Developers get something today. Google says customers can connect local Codex CLI and Desktop workflows to its Gemini platform for inference starting October 21. OpenAI models are also coming to Gemini for business and Google Antigravity, with no date.

Google now claims to be "the only hyperscale cloud provider offering access to all three leading closed source LLMs."

The procurement wall just came down

If you work inside a company on Google Cloud, you know how model choices get made. Often it is not the benchmark. It is which vendor is already approved, which contract has money left in it, and which data agreement legal already signed.

That used to steer Google shops toward Gemini and Claude, which Google already offered. Now GPT-6 sits in the same place, under the same security review and the same bill.

For OpenAI, it is reach into budgets it could not touch before. For Google, it is a bet that you would rather buy every model from one cloud than leave for the one model you want. Box is an early example: its VP of AI Products, Yashodha Bhavnani, said in Google's post that Box "will be able to give customers access to OpenAI's frontier AI models alongside Google Cloud's security and compliance capabilities."

Yes, this is a press release about a preview. Nobody outside a selected group can run it yet, and nobody knows the price.

But the direction is clear. With OpenAI, Google and Anthropic models on one cloud, the deciding question moves from "can we buy it" to "which model fits this task."

What it changes for your team

If you are already a Google Cloud customer:

• Fill in Google's interest form if you want GPT-6 in the preview.
• Ask your account team how OpenAI spend counts against your existing commitment.
• Hold sensitive workloads until Zero Data Retention ships, if your policy requires it.

If you run Codex locally, check whether routing inference through Google's platform fits your billing better than paying OpenAI directly. That option exists now.

If you build on a gateway that routes between providers, this is one more route. Run your own evals before you switch anything.

What to watch

Two things will decide whether this matters: the per-token price on Google Cloud compared with OpenAI's own API, and the date it leaves private preview.

Watch the price sheet. Then redo your math.