Google Cloud began rolling out three OpenAI models to selected enterprise customers on 21 October 2026. Its Model Garden announcement identifies GPT-Astra, GPT-6.1 Sol and GPT-6 Luna as the first models in private preview. OpenAI says general availability is planned in the coming weeks. For now, this is a limited preview, not a launch for every Google Cloud customer.
The practical change is where participating teams can evaluate and purchase the models. Google says preview customers can use its APIs, while model purchases will be transacted through Google Cloud Marketplace. That could let an organization already using Google Cloud handle model spending within an existing cloud procurement process. The applicable contract and preview access still need checking before a team plans a migration.
Access and purchasing
Google says OpenAI model usage will appear on a Google Cloud invoice and that eligible customers can apply existing Google Cloud spending commitments. Those are the announced commercial routes, not a published price comparison. The companies have not supplied performance or cost results for a like-for-like workload in these announcements.
Google also says selected customers can connect local Codex CLI and Desktop workflows to its platform for inference. OpenAI describes Codex workflows as part of the partnership. Neither announcement gives enough implementation detail to treat every Codex workflow as generally available through Google Cloud.
Governance claims need workload-level checks
Google describes access controls, data perimeters, isolated infrastructure and unified auditing for the offering. It says customer prompts, context and outputs remain within a designated perimeter and are not used to train foundation models. These are Google’s statements about its service, not an independent assessment of a customer’s deployment.
Two qualifications matter: Google labels Zero Data Retention options and Confidential Compute support as coming soon. Teams with strict data-handling requirements should confirm the preview’s actual regions, retention terms, identity controls and data flow before putting sensitive workloads on it.
What remains ahead
OpenAI expects general availability in the coming weeks. The companies also describe later availability in Google Antigravity and other Google enterprise products; the announcements do not establish a common release date for those integrations. The immediate decision for an operator is whether the limited Model Garden preview merits an evaluation under its current access and governance terms, rather than assuming the full proposed integration is already in production.
