Google Cloud is now offering OpenAI models through Model Garden. GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna are available in private preview for select enterprise customers. This makes Google the only hyperscaler to provide access to the three most powerful closed-source LLMs, which, in addition to GPT-6 Astra, include Claude Fable 5.1 and Google’s own Gemini 4 Argon.
OpenAI “grew up” within Azure, where GPT models were initially exclusive due to an earlier investment by Microsoft. In the summer of 2025, more than two years after the ChatGPT hype began, AWS followed suit. Now comes this notable collaboration with Google, a direct LLM competitor.
Model Garden
Google Cloud is bringing OpenAI’s latest models to its so-called Model Garden, where more than 200 proprietary, third-party, and open-source models are already available. In addition to Gemini, Google has been offering models from Anthropic for quite some time. With the arrival of GPT, Google is now the only hyperscaler with access to all three leading closed-source LLMs. This is, incidentally, a deliberate choice by Google, as it has opted to keep Gemini exclusively on its own cloud service.
Previously, the relationship between the two companies focused on providing capacity for OpenAI, without the models being included in the broader pool of LLM options available to Google Cloud customers. In 2025, OpenAI and Google signed a deal for additional computing capacity, after which Google was officially named OpenAI’s cloud partner.
Three GPT-6 variants
Initially, there are three models. GPT-6 Astra is designed for heavy-duty tasks such as coding, computer use, and scientific research. GPT-6.1 Sol approaches Astra in terms of intelligence but does so at a lower cost. GPT-6 Luna is the fast, low-cost model for everyday work at scale, as OpenAI describes it.
“Modern enterprise AI requires that each workload be matched to the right model, rather than relying on a single model for every task,” said Michael Gerstenhaber, VP of Product Management at Google Cloud. Because sales are handled through the Google Cloud Marketplace, all usage appears on a single invoice and counts toward existing cloud commitments and Flexible Savings Plans.
Data remains with Google
Google is also making specific commitments regarding governance now that the OpenAI models are among the fully supported LLMs. Prompts, context, and output remain within the customer’s data perimeter and are not used to train models. This is a guarantee that previously could not be made outside of the Model Garden. The OpenAI models run on Google Cloud infrastructure, and queries pass through isolated gateways. Zero Data Retention and Confidential Compute will follow shortly and apparently require some additional work.
For now, the rollout is limited to a private preview. However, developers can already connect their Codex CLI and Desktop workflows to the Gemini platform for inference. Support for Gemini for Business and Google Antigravity will follow later.