Google Cloud Lists Three OpenAI GPT-6 Models in Private Preview, With No Price and No Release Date
A.I. / news
Google Cloud Lists Three OpenAI GPT-6 Models in Private Preview, With No Price and No Release Date
Codex CLI can point at Google's platform today. Zero Data Retention and Confidential Compute are still listed as not yet available.
Google Cloud began offering three OpenAI models to select enterprise customers on October 21, in a private preview that is not generally available, according to a Google Cloud blog post by Michael Gerstenhaber, VP of Product Management, Gemini, Google Cloud. The post names no per-token price and no date for wider release.
OpenAI's own news feed carries a matching item, saying OpenAI and Google Cloud are bringing frontier models and Codex workflows to Model Garden through Google Cloud Marketplace. Its full page could not be retrieved, so the details below come from Google's post and from Techzine Global.
Three GPT-6 models in Model Garden
Google lists GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. It describes Astra as the model for demanding coding, computer use, professional and scientific work.
Sol is described as near-Astra intelligence at lower cost. Luna is the fast, cheaper option for everyday work at scale.
Interested customers apply through a form on cloud.google.com. Google says it is now the only hyperscale provider offering Gemini, Claude and GPT, alongside more than 200 other models.
What works on Oct. 21 and what does not
Starting October 21, customers can connect local Codex CLI and Desktop workflows to Google's Gemini platform for inference, the post says. Several enterprise controls are not yet live.
| Item | Status in Google's post |
|---|---|
| GPT-6 Astra, 6.1 Sol, 6 Luna | Private preview, select customers |
| Codex CLI and Desktop inference | Available from Oct. 21 |
| Zero Data Retention options | Not yet available |
| Confidential Compute | Not yet available |
| OpenAI models in Gemini for business and Antigravity | Not yet available |
Techzine Global reported that the two controls still need additional work. Google's post says prompts, context and outputs stay inside the customer's data perimeter and are never used to train foundation models.
Billing through Google Cloud Marketplace
OpenAI models are transacted through Google Cloud Marketplace, per the post. Usage appears on one Google Cloud invoice and draws down existing cloud commitments.
Spend also counts toward Flexible Savings Plans across model providers. The post does not say what OpenAI charges per token or whether Marketplace adds a margin.
Yashodha Bhavnani, VP of AI Products at Box, said in the post that Box will use OpenAI models through Google Cloud to give customers frontier models alongside Google's security and compliance. Box is the only customer quoted.
How the OpenAI and Google relationship got here
Techzine Global wrote that OpenAI's models were first exclusive to Azure because of Microsoft's investment in OpenAI. It said AWS began offering them in summer 2025, and that the two companies signed a compute deal in 2025 that made Google OpenAI's cloud partner.
Until now that relationship ran one way, with Google supplying capacity to OpenAI rather than selling its models to Google Cloud customers, Techzine said.
Related reporting on this site includes Anthropic's Claude Duet waitlist and the RuneBench agent benchmark, where GPT-6 Astra ranked first at 7.3 percent on one aggregator's list.
What is still unknown
The post does not list supported regions or data residency beyond the perimeter language. It gives no general availability date and no Codex setup steps.
Google lists Gemini for business and Google Antigravity support, and further Model Garden additions, as upcoming. It has not given dates.
Sources
More in A.I.
- 01OpenAI Says AI Found the PHP Flaws Before Attackers. Cloudflare Says They Leaked and Drew Attack TrafficThe two Oct. 21 accounts of the php2shell disclosure measure different things. Trail of Bits' Patch the Planet dashboard shows 466 of 1,995 reported issues patched.
- 02Meta Releases Muse Spark Open, an 816B Model, With a $20 Million Revenue Trigger in the LicenseThe weights are on Hugging Face behind a sign-up form. Meta's own benchmark table has it trailing rivals on coding, and one outlet misreports the licence threshold.
- 03Flapping Airplanes Raises $1 Billion at a Reported $6 Billion Valuation Nine Months After Its SeedThe lab's pitch is models that need far less training data. Its only technical detail so far is a GPU megakernel meant to keep chips busy.
- 04Trail of Bits Finds 21 Regressions in 130-Plus Merged Patches With a Codex Plugin, and None Touched SecurityThe security firm's regression-hunt plugin cut false positives 25% against a baseline Codex prompt, while the Patch the Planet dashboard shows 466 of 1,995 reported issues patched upstream.