OpenAI and Google Cloud announced that GPT-6 Astra, GPT-6 Sol, GPT-6 Luna and the Codex workflow will land in Model Garden within weeks, first in private preview. Enterprises will be able to pay through existing Google Cloud Marketplace spending commitments, completing OpenAI's presence across all three major clouds.
OpenAI's models are heading to Google Cloud, and enterprise procurement just got another aisle
OpenAI and Google Cloud announced a partnership on Wednesday, October 21 that will bring OpenAI's frontier models and its Codex agentic coding workflow to Google Cloud Model Garden, purchasable through Google Cloud Marketplace. The first batch includes three models: GPT-6 Astra, the flagship aimed at coding, computer use, professional work and scientific research, plus GPT-6 Sol and GPT-6 Luna for lighter workloads. Availability is promised within weeks, starting with a private preview.
The commercial detail is the part enterprise buyers will care about most. Customers can pay for OpenAI models through their existing cloud spending commitments on Google Cloud Marketplace, no separate billing relationship with OpenAI required. For companies that have already pledged large sums to Google, that turns a procurement headache into a line item.
Brockman's framing, and what it signals
Greg Brockman, OpenAI's president, framed the move as an important step toward putting the most capable models and Codex where customers are already building. The line summarizes OpenAI's distribution strategy: be present in every environment where developers already live, rather than forcing them out of it.
That strategy has been unfolding all year. OpenAI's models already run on Microsoft Azure, its founding partner and largest investor, and on AWS Bedrock since the spring. Google Cloud becomes the third leg. For a company whose fortune was built on an exclusive Microsoft relationship, the multi-cloud pivot is now complete.
Rivals deepening ties, again
The awkward part, of course, is that OpenAI and Google compete directly for chatbot users. Gemini and ChatGPT fight for the same audience. Yet the two companies are bound together by cloud and search agreements that entangle their economics. Selling your rival's models on your own platform no longer registers as an anomaly in this industry. It's just how the business works now.
Google, for its part, gets to present Model Garden as the neutral catalog where Gemini, open-weight models and now GPT-6 models coexist. Google's own announcement, credited to Gemini product VP Michael Gerstenhaber, leaned into that framing. A cloud that sells everything is harder to argue against than a cloud that sells only itself.
What to watch in the coming weeks
The private preview will give way to broader availability, and no timeline has been shared. For enterprises, the practical effect is bargaining power: the same model on three platforms invites comparison on price, latency and data governance, and weakens the lock-in each cloud has enjoyed. Why settle for one vendor's terms when the same weights are on three shelves? For OpenAI, every new channel reduces dependence on any single partner and widens the funnel into corporate accounts. For Google, hosting Gemini's chief rival is a calculated trade, revenue and relevance over exclusivity. The only certainty is that the next round of enterprise cloud contract negotiations just got more interesting for buyers.
