OpenAI Models Google Cloud Private Preview Begins

Here is a sentence you would not have believed a year ago. Google Cloud now sells access to OpenAI models, the same family that powers ChatGPT. On October 21, 2026, Google Cloud announced that GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna are rolling out in private preview in its Model Garden. Google Cloud’s launch of the OpenAI models matters because it makes Google Cloud, in Google’s own words, the only hyperscale cloud with all three leading closed models on one platform. If your company already pays a Google Cloud bill, this changes how you buy AI, how you route Codex workflows, and how you think about model choice for the next year.

How OpenAI Models Reached Google Cloud

Google Cloud published its announcement on October 21, 2026, written by Google Cloud’s Michael Gerstenhaber. The post says the three models are rolling out in private preview to select enterprise customers, with more coming soon. OpenAI’s matching post said general availability is planned in the coming weeks and added one notable line. Microsoft remains OpenAI’s primary cloud partner.

This is the third big cloud route to GPT-6, not the first. Microsoft Foundry made GPT-6 Astra generally available on September 3, 2026, and Amazon Bedrock followed on September 8, with both also listing Sol and Luna. On October 21, Google’s docs did not yet list any GPT-6 model, so the announcement is ahead of the catalogue.

Why Google Cloud Cares About Model Choice

The announcement rests on one idea. Enterprise AI requires matching each workload to the right model, not using one model for everything. Google says it lets teams balance cost, capability, and performance. The three models cover different tiers. GPT-6.1 Sol offers near-Astra intelligence at lower cost for agentic coding, computer use, and professional work. GPT-6 Luna is a fast, cost-efficient model for everyday work at scale.

Google also claims it is the only hyperscale cloud offering all three leading closed LLMs. That is Google’s own wording. Google is the only one of the three whose platform also includes Gemini, and the only one where GPT-6 is still in private preview rather than generally available.

Billing, Data and Security for OpenAI Models on Google Cloud

Billing is the most concrete part of the announcement. OpenAI models are transacted through Google Cloud Marketplace, so usage appears on a single Google Cloud invoice, draws down existing commitments, and counts toward Flexible Savings Plans. There is no separate OpenAI contract. Third-party products get no discounts from the Flexible Savings Plan, so the benefit is commitment drawdown, not a cheaper token.

What the token costs on the Google route is unpublished as of October 21. OpenAI’s own API list prices, cited as company figures, put Astra at 10 and 50 dollars per 1 million input and output tokens, Sol at 2 and 10, and Luna at 10 and 50 cents. On security, prompts and outputs stay inside the customer’s data perimeter and are never used to train models. Zero data retention and Confidential Compute are both listed as coming soon.

What Is Still Missing Before Launch

A private preview means the announcement is a promise, not a product you can switch on today. Several things a buyer would ask for are absent. No published price per token on the Google route. No named regions. No model identifiers or quotas. No GA date beyond OpenAI’s coming weeks. And the earlier GPT-6 Sol, which AWS and Azure still sell, is not named.

If your policy requires zero data retention for any third-party model, the Google route does not meet it yet, while AWS documents ZDR on request today. One use case looks better. Google says customers can connect their local Codex CLI and Desktop workflows directly to the Gemini platform for inference starting now. For background, see our OpenAI GPT news explained guide and our Gemini 4 Argon features rundown.

What to Watch on Google Cloud Next

The next milestones are easy to name. Watch for the public catalogue entry, the first published rate card, named regions, and the zero data retention launch. Google also says the models are coming to Gemini for business and Google Antigravity. Teams already running GPT-6 on Azure or AWS with a Google commitment to spare should probably wait for the rate card and compare against today’s price, since moving for drawdown alone can cost more per token.

Teams that mainly want an evaluation path should submit the interest form now and test under the preview without moving production workloads. The Unite.AI coverage, the Digital Applied analysis, and the Blockchain. News report all dig into the gaps above.

Conclusion

Google Cloud now offers GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna in private preview, billed through one Marketplace invoice. It is the third cloud to sell GPT-6 and the only one that also sells Gemini. Pricing, regions, and zero data retention are still unpublished, so treat this as a preview to evaluate. If you are a Google Cloud shop, your one useful next step is to submit the interest form and evaluate the tier that fits your workload before general availability lands.