OpenAI’s models are finally fully supported on Google Cloud
Google Cloud now offers OpenAI models through Model Garden. GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna are available in a private preview for selected business customers. This makes Google the only major cloud provider to offer access to the three leading closed-source large language models (LLMs): OpenAI’s GPT-6 Astra, Anthropic’s Claude Fable 5.1, and Google’s own Gemini 4 Argon.
OpenAI first grew within Azure, where GPT models were initially available only through Microsoft’s cloud because Microsoft had invested in the company. In the summer of 2025, more than two years after ChatGPT became popular, AWS also began offering access. Now, OpenAI is working with Google, a direct competitor in the LLM market.
Model Garden
Google Cloud is adding OpenAI’s newest models to Model Garden. This collection already contains more than 200 models from Google, other companies, and open-source projects. Google has also offered models from Anthropic for some time, alongside its own Gemini models. With GPT models now available, Google is the only major cloud provider that offers access to all three leading closed-source LLMs. Google has made a deliberate choice to keep Gemini exclusive to its own cloud service.
Until now, the two companies mainly worked together to give OpenAI more computing power. Google Cloud customers could not choose OpenAI’s models from the wider range of LLMs available on the platform. In 2025, OpenAI and Google signed a deal for extra computing power. After this, Google was officially named one of OpenAI’s cloud partners.
Three GPT-6 variants
At first, three models will be available. GPT-6 Astra is designed for demanding tasks such as writing code, using computers, and doing scientific research. GPT-6.1 Sol offers similar intelligence to Astra but costs less. GPT-6 Luna is a fast, low-cost model for everyday tasks that need to be handled on a large scale, according to OpenAI.
“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. Customers buy the models through Google Cloud Marketplace. This means all usage appears on one bill and counts toward their existing cloud spending commitments and Flexible Savings Plans.
Data stays with Google
Google has also made specific promises about how customer data is protected now that OpenAI’s models are fully supported on its platform. Customer prompts, context, and outputs stay within the customer’s data boundaries and are not used to train models. Google could not make this same guarantee for models outside Model Garden. OpenAI’s models run on Google Cloud’s systems, and requests pass through separate, isolated gateways. Zero Data Retention and Confidential Compute will be added soon, although some more work appears to be needed first.
For now, access is limited to a private preview. However, developers can already connect their Codex CLI and Desktop workflows to the Gemini platform to run models and receive results. Support for Gemini for Business and Google Antigravity will come later.