OpenAI’s frontier models are now for sale inside Google Cloud, and customers can pay for them out of money they have already committed to Google. The launch on October 21 was a private preview of three models. The bigger change is that OpenAI models will run inside Google’s own agent products, which turns a rival platform into an OpenAI sales channel.
OpenAI’s announcement says the two companies “are partnering to bring OpenAI’s frontier models and Codex agentic workflows into Google Cloud’s Model Garden.” Google’s companion post names the models: “OpenAI’s GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna models are rolling out now in private preview to select enterprise customers in Google Cloud’s model garden, with more coming soon.” GPT-6.1 Terra, released five days later, is not on the list.
Paid for with Google commitments
The billing terms matter more than the catalog entry. “Google Cloud customers will be able to purchase OpenAI frontier models via model garden, transacted through the Google Cloud Marketplace,” OpenAI writes. “Enterprises with Google Cloud spend commitments will be able to draw down their GCP commitments for these purchases.” Google adds that usage “counts toward Flexible Savings Plans.”
Large companies sign multi-year spending commitments with their cloud providers, and procurement teams treat that money as already allocated. Letting customers spend it on OpenAI models removes the usual obstacle of a new vendor contract. It also puts OpenAI in reach of budgets it previously had to win separately, with the transaction running through Google’s marketplace. Alphabet’s third-quarter release a week later put Google Cloud revenue at $30.7 billion, up 103%, and on the earnings call, as transcribed by Investing.com, the company put Google Cloud’s backlog at $561 billion. That is the pool of committed spending OpenAI can now draw from.
Google Cloud chief Thomas Kurian described the logic as choice. “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,” he told SiliconANGLE.
Inside Google’s agent products
The detail with the longest reach is in OpenAI’s post: “OpenAI models will soon be available in Google Antigravity and the Gemini Agent.” Both are Google products built around Gemini, and both will soon offer a competitor’s models.
Codex moves in the other direction. Google says customers can “connect their local Codex CLI and Desktop workflows directly to our Gemini platform for inference.” OpenAI’s coding agent, running on a developer’s machine, can send its model calls through Google’s infrastructure. Google says “OpenAI models run on dedicated Google Cloud infrastructure.”
On Hacker News, where the thread reached 326 points, one commenter quoted the Gemini Agent line and noted Google would “need to have a better name if they’re going to use 3p models.”
What is not there yet
The launch is a preview with several pieces still missing. General availability “is planned in the coming weeks.” Google lists both “Confidential Compute coming soon” and “Zero Data Retention (ZDR) options available coming soon,” the two controls regulated buyers usually require before production use. Neither post names regions or per-model pricing. TNGlobal noted that “The companies have not disclosed customer uptake, regional rollout dates or pricing for each model in this announcement.”
OpenAI ended its post with a sentence aimed at a third party: “Microsoft remains OpenAI’s primary cloud partner.” Another Hacker News commenter picked up on it: “Gotta love the disclaimer at the bottom.” Microsoft made GPT-6.1 Terra generally available in Foundry the day OpenAI launched it. Google’s preview covers three models, not including the newest, and lacks the controls regulated buyers need. For now, being the primary partner means getting new models first and fully supported.
The longer-term shift is that frontier labs are selling through every cloud that holds customer commitments, including a cloud run by their most direct model competitor. Inside Antigravity and the Gemini Agent, Gemini and GPT will compete on the same terms. Which one customers pick there will show how much model choice matters once distribution is no longer the deciding factor.