OpenAI puts GPT-6 Astra, Sol and Luna, and Codex, into Google Cloud’s Model Garden
The companies, Wednesday: GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are in private preview for selected enterprise customers, with general availability planned in the coming weeks. Codex’s command-line and desktop apps can call the models for inference now. Google Antigravity comes later. OpenAI says Microsoft remains its primary cloud partner. Neither post states a price.
Deep dive4 minIn-depth
Maya Chen4 min read
- OpenAI
- Google Cloud
- Codex
- GPT-6
- Model Garden

Photo: Google CloudSource page
OpenAI and Google said on Wednesday that OpenAI’s frontier models and its Codex coding agent are coming to Google Cloud’s Model Garden. Google’s post, by Michael Gerstenhaber, says GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are rolling out now in private preview to selected enterprise customers. OpenAI says general availability is planned in the coming weeks, and that customers can request access. The last line of OpenAI’s post is that Microsoft remains its primary cloud partner.
Google describes the three models by the job it wants each one sold for. Astra is for demanding work across coding, computer use, professional tasks and scientific research. Sol is “near-Astra intelligence at lower cost” for agentic coding, computer use and professional work. Luna is “a fast, cost-efficient model for everyday work at scale.” Those are Google’s labels. Wednesday’s posts do not include a benchmark table, a token price, or a date inside the “coming weeks.”
One invoice, and Codex on the desk
Customers buy the models through Google Cloud Marketplace. Google says the usage appears on a single Google Cloud invoice, draws down existing cloud commitments, and counts toward Flexible Savings Plans across model providers. Starting Wednesday, Google says, customers can call the models through its APIs and can connect a local Codex command-line or desktop app to what it calls the Gemini platform for inference. OpenAI’s wording on the agent is that Codex workflows come into Model Garden. The two sentences are not identical. One is a local coding app pointed at Google for inference. The other is Codex as a product inside the catalogue.
A second destination is explicitly later. OpenAI says the models will soon be available in Google Antigravity and in the Gemini Agent. Google’s post says they are coming soon to Gemini for business and to Google Antigravity. The posts do not say whether “the Gemini Agent” and “Gemini for business” are the same product. They also do not say what Antigravity is, beyond the place developers will get a choice of model.
Google’s security claims are specific, and two of them are dated. Prompts, context and outputs stay inside the customer’s data perimeter and are never used to train foundation models. Zero Data Retention is “coming soon.” The models run on dedicated Google Cloud infrastructure, with Confidential Compute “coming soon.” Queries go through what Google calls hardened, isolated gateways, and Google says third-party model providers have no unauthorized access to the customer’s cloud environment. Every agent and model call sits under an authenticated identity, with VPC Service Controls and unified audit logs. Those are the vendor’s assurances. The post does not name an auditor.
This partnership is an important step in making our most capable models and Codex available where customers are already building, giving them even more choice in how they use frontier AI as we grow the agentic ecosystem together.
— Greg Brockman, OpenAI president and co-founder, Oct 21, 2026
We are focused on giving customers a broad selection of models, including those from OpenAI, so they can choose the right model for any given job, helping to deliver optimal quality and lower costs.
— Thomas Kurian, chief executive of Google Cloud, Oct 21, 2026
The third large cloud
Krystal Hu, reporting for Reuters from San Francisco, wrote that Alphabet is the latest major hyperscaler to offer OpenAI’s models, after OpenAI’s exclusive distribution arrangement with Microsoft expired in April. OpenAI has since expanded distribution to Amazon Web Services and now to Google, she wrote. The Next Web added a date Reuters’ story did not: OpenAI put its agents on Amazon Bedrock last month. Google, in its own post, now calls itself the only hyperscale cloud with all three leading closed models, next to more than 200 first-party, third-party and open models. Claude is already on that platform. Gemini is Google’s.
Hu also wrote that cloud providers usually take a cut of sales through their platforms. She reported that an IPO filing Reuters reviewed shows 47 percent of Anthropic’s 2025 revenue came through Amazon’s and Google’s clouds. That figure is Reuters’ reading of a filing it says it reviewed. It is not in Wednesday’s posts, and it is not a number for this deal. Neither company said what share Google will keep, or what a customer will pay per million tokens. Hu noted that Google and OpenAI still compete in models and in search, and that Google had already agreed last year to offer OpenAI more computing capacity. Kurian has told Reuters, in an earlier interview she cites, that the platform is open to any model popular with developers.
What a buyer can do on Wednesday is narrower than the headline. A selected enterprise customer can ask for the private preview, point Codex at Google for inference, and, if accepted, pay through the cloud bill they already have. General availability, Antigravity, a zero-retention option and confidential computing are all still ahead of them, on the companies’ own timing. Microsoft is still the partner OpenAI names first.


