OpenAI's GPT-6 Models Reach Google Cloud's Model Garden in Private Preview, Techzine Reports

OpenAI and Google Cloud announced on October 21 that OpenAI's frontier models will be sold through Google Cloud's Model Garden, after Microsoft Azure and AWS. Techzine reports GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are in private preview for select enterprise customers, with no general availability date.

9to5Google reports something looser. Its piece says the models arrive "in the coming weeks" and that neither company has confirmed which GPT models are included. OpenAI's own post returned an access error when fetched for this report, so the two accounts could not be reconciled against it.

What Techzine says is in the preview

Per Techzine, the three models split by price and speed.

Model	Positioning in the announcement
GPT-6 Astra	Demanding tasks: coding, computer use, scientific research
GPT-6.1 Sol	Close to Astra in capability at lower cost
GPT-6 Luna	Fast, low-cost model for everyday work at scale

No prices or rate limits appear in either report. Customers buy through Google Cloud Marketplace, and usage appears on one invoice that counts toward existing cloud commitments and Flexible Savings Plans.

Developers can already connect Codex CLI and Desktop workflows for inference, Techzine says. Support for Gemini for Business and Google Antigravity is planned for later, and 9to5Google says Antigravity access is coming "soon".

What is still missing

Zero Data Retention and Confidential Compute will follow, but Techzine says each needs more work first. Google said prompts, context and output stay within the customer's data perimeter and are not used to train models, and that queries pass through isolated gateways on Google Cloud infrastructure.

For regulated buyers, the missing retention option matters more than the missing date. Public-sector buyers are the cautious end of the market: Swedish government offices have already blocked ChatGPT, Claude and Copilot, and a data perimeter is a weaker promise than zero retention.

Why Google wants three closed models

Thomas Kurian, CEO of Google Cloud, said in an excerpt quoted by 9to5Google: "We are focused on giving customers a broad selection of models, including those from OpenAI." Michael Gerstenhaber, VP of Product Management at Google Cloud, said enterprise AI should match each workload to the right model "rather than relying on a single model for every task," per Techzine.

Techzine reports that Google now says it is the only hyperscaler offering access to three leading closed-source models: GPT-6 Astra, Claude Fable 5.1 and Google's own Gemini 4 Argon. That is Google's claim, relayed by Techzine, and it depends on the preview counting as access. Model Garden lists more than 200 proprietary, third-party and open-source models.

The history behind the deal

OpenAI's models were exclusive to Microsoft Azure at first because of Microsoft's earlier investment, Techzine says. AWS followed in summer 2025. Until now Google's relationship with OpenAI covered only compute capacity, under a 2025 deal that named Google an OpenAI cloud partner.

The arrangement runs one way. OpenAI models come to Google Cloud, and Google keeps Gemini exclusive to its own cloud, Techzine says. OpenAI itself has been reworking how it handles its most capable systems, including moving frontier training offline.

OpenAI and Google have not published a general availability date. 9to5Google puts the wider rollout at "the coming weeks" from October 21.