In this article
  1. What happened
  2. Why it matters for builders and small teams
  3. What to do
  4. What we’ll watch

TL;DR: On October 21, 2026, OpenAI and Google Cloud announced that OpenAI’s models and Codex are coming to Google Cloud’s Model Garden. Google says GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are rolling out now in private preview “to select enterprise customers,” and OpenAI says the preview is invite-only, with general availability “planned in the coming weeks.” Usage is bought through Google Cloud Marketplace, so it lands on your Google Cloud invoice and can draw down existing Google Cloud spend commitments. Neither company has published prices. If your business already runs on Google Cloud, register interest now; otherwise, nothing changes for you today.

What happened

OpenAI published its announcement on October 21 at 12:00 UTC (8:00 p.m. Manila), and Google Cloud posted its own the same day. According to the two companies:

  • The models: Google lists three OpenAI models in Model Garden. GPT-6 Astra is “designed for demanding work across coding, computer use, professional tasks, and scientific research.” GPT-6.1 Sol offers “near-Astra intelligence at lower cost,” and GPT-6 Luna is “a fast, cost-efficient model for everyday work at scale.”
  • Who can use them now: Google says the models are “rolling out now in private preview to select enterprise customers.” OpenAI’s sign-up page says the preview is “invite-only” and that select customers can evaluate the models through the Responses API on Google Cloud “for supported development and internal testing use cases.”
  • Codex: Google says customers can “connect their local Codex CLI and Desktop workflows directly to our Gemini platform for inference.” OpenAI’s page lists Codex through the API in the Codex app, the CLI and supported IDEs as part of the limited preview.
  • Billing: Google says usage “appears on a single Google Cloud invoice, draws down existing cloud commitments, and counts toward Flexible Savings Plans (FSP) across model providers.” OpenAI says enterprises “with Google Cloud spend commitments will be able to draw down their GCP commitments for these purchases.”
  • Data: Google says prompts, context and outputs “remain entirely within your designated data perimeter and are never used to train foundation models.” It says the models run on dedicated Google Cloud infrastructure, and that Zero Data Retention options and Confidential Compute are “coming soon.”
  • Next: both companies say OpenAI models will come to Google Antigravity, and to what Google calls Gemini for business and OpenAI calls the Gemini Agent. Neither gives a date.
  • Microsoft: OpenAI says, “Microsoft remains OpenAI’s primary cloud partner.”

Why it matters for builders and small teams

For teams that already run on Google Cloud, this puts OpenAI’s models in the same console, billing account and access controls as Gemini and the other Model Garden models, instead of a separate vendor account and bill. Google says it is now the only big cloud provider offering “all three leading closed source LLMs,” though it doesn’t name the three.

For a small team, the practical gains are fewer vendor accounts, one invoice, and the option to spend Google Cloud commitments you have already signed on OpenAI models. The catch is timing: today it’s an invite-only preview for select enterprise customers, and there are no published prices to compare with OpenAI’s own API. In our view, a small business without a Google Cloud commitment has no reason to switch anything yet.

What to do

  1. If you’re on Google Cloud, register interest. Google and OpenAI each link a form from their announcements. Ask your Google account team whether your billing account qualifies for the preview.
  2. Check what your commitment covers. If you have a committed-spend deal or Flexible Savings Plan, ask Google how Marketplace purchases of OpenAI models count against it before you plan around it.
  3. Don’t move production yet. OpenAI limits the preview to development and internal testing use cases, and Zero Data Retention isn’t available yet. Keep production traffic on the API you use today until general availability and pricing are public.
  4. Plan for model switching. If you use Codex or build agents, keep model names and endpoints in config rather than hard-coded, so you can test the Google Cloud route side by side when you get access. Our story on OpenAI’s 12-hour prompt caching covers the cost levers on OpenAI’s own API, and our Antigravity model changes story covers which models Antigravity offers today.

What we’ll watch

We’ll update this story when OpenAI or Google publishes prices, regions or a general availability date, or when OpenAI models appear in Antigravity.

Sources & method

Sources

Method. Last checked October 21, 2026, at 10:07 p.m. Manila time (14:07 UTC). Model names and descriptions, availability, billing, data and security terms, and the quotes come from the two announcements and OpenAI’s sign-up page. The announcement time comes from OpenAI’s news feed. Neither company published prices or regions. The advice in “Why it matters” and “What to do” is our own judgment, and we haven’t had access to the preview. No affiliate links on this page. How we work.

Spotted an error? See our corrections policy and tell us.