Google Cloud is adding OpenAI’s latest model family to its Model Garden, giving a limited group of enterprise customers another way to run generative AI workloads inside Google’s cloud environment. The company said GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are rolling out in private preview to select customers, rather than being generally available.
The move broadens the model choices Google Cloud sells alongside its own Gemini models and other third-party systems. Google described Astra as the highest-capability option of the three for demanding coding, computer-use, professional and research tasks. It positioned Sol as a lower-cost model with near-Astra performance, while Luna is intended for faster, high-volume everyday work. Those descriptions are the vendor’s characterization and should not be read as independent performance testing.
How access and billing will work
Customers in the preview can use the OpenAI models through Google Cloud APIs and connect local Codex CLI or Desktop workflows to the Gemini platform for inference. Google said use will be transacted through Google Cloud Marketplace, appear on the customer’s existing cloud invoice and count toward eligible Flexible Savings Plans. That arrangement is designed to reduce the need for separate vendor contracts and billing systems.
Google also said developers can select and swap models based on capability, latency and cost within Model Garden. The company calls itself the only hyperscale cloud provider offering access to the three leading closed-source model families on one enterprise AI platform, but that remains a competitive claim from Google rather than a neutral market assessment.
Security promises and availability limits
Google says prompts, context and outputs remain inside the customer’s designated data perimeter and are not used to train foundation models. It also lists authenticated identities, audit logging and VPC Service Controls as available governance features. Two other controls are not yet available: Zero Data Retention options and Confidential Compute support are both described as coming soon.
The distinction matters because private preview access is narrower than a public preview or a general release. Google has not announced broad availability dates or public pricing for the three models. It said more access is coming and plans to add OpenAI models to Gemini for business and Google Antigravity, but those integrations are also future commitments.
Why it matters
For enterprise buyers, the announcement is less about a single model launch than about how cloud platforms package competing AI systems. Centralized identity, network controls, procurement and logging can be as important as model performance when regulated or sensitive workloads move from experiments into production. Google’s approach also puts model choice inside the same commercial framework as other cloud spending, which could make adoption easier for existing customers while increasing their dependence on the platform’s governance layer.
Source
Google Cloud: Google Cloud expands model choice and flexibility, bringing OpenAI models to customers.
Featured image: AI-generated editorial illustration; it does not depict an actual event.
