OpenAI Models Head to Google Cloud in Private Preview

OpenAI and Google Cloud announced a partnership on October 21 to bring OpenAI's frontier models and Codex agentic workflows to Google Cloud's Model Garden. The companies describe a route for customers to use those capabilities within their existing cloud environment.

Google says GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are rolling out in private preview to selected enterprise customers. OpenAI says general availability is planned in the coming weeks. This is therefore not unrestricted access for every customer today.

The announcement includes purchasing through Google Cloud Marketplace and the ability to draw down eligible Google Cloud spending commitments. The companies also describe access controls and governance features intended for enterprise use.

Support in Google Antigravity and Google's Gemini business offering is described as coming soon. OpenAI states that Microsoft remains its primary cloud partner.

For businesses, the useful change is another procurement and deployment option. An additional purchasing route does not automatically make an existing application cheaper, faster or more accurate.

JetherVerse's practical takeaway is to compare one real workflow before changing providers or infrastructure. Use the same examples, success criteria and failure cases across the options being evaluated.

For example, a support assistant should be tested on questions with incomplete records as well as straightforward requests. Check whether it asks for missing information, uses the right source and avoids inventing a confident answer.

The comparison should include the full operating arrangement. Record usage charges, access requirements, logging, retention settings and the effort needed to maintain the integration. Verify which features are available in the chosen release rather than assuming that planned controls are already active.

Keep the existing route available during a pilot. That makes it easier to investigate unexpected behaviour without turning a procurement experiment into an interruption for customers.

For teams deciding where an AI feature should run, those requirements and evaluation checks fit naturally into JetherVerse's AI solutions work: https://jetherverse.com/services/ai-solutions