Google Cloud is now bringing in the models from OpenAI into its Model Garden, allowing enterprise customers to use the OpenAI platform alongside Google’s Gemini models and other artificial intelligence (AI) systems. The first release includes GPT-Astra, GPT-6.1 Sol and GPT-6 Luna, which are available in private preview to a limited set of customers.
The announcement by Michael Gerstenhaber, vise president of product management for Gemini at Google Cloud, expands a catalog that includes more than 200 first-party, third-party and open models. Google’s cloud platform now gives businesses the ability to select models that fit their workload needs by offering access to all three of the most prominent proprietary large language model families.
3 OpenAI Business Models for Different Business Needs
The three models have different goals. GPT-Astra implements difficult projects in the areas of scientific research, professional work, computer interaction and software development. GPT-6.1 Sol is targeted at coding and business applications that require performance similar to that of Astra’s, but at a lower price. GPT-6 Luna aims to efficiently and quickly manage daily tasks.
Businesses can choose models according to the complexity of the task, the response time and the operating expenses. For instance, a company developing an AI coding assistant could use a cheaper model for routine requests, and save Astra for more complex programming challenges.
Google also wants to make it easier to switch between supported models, so you don’t need to have separate vendor contracts. There may be differences in the model behavior , tool support and response formats , so developers will still have to judge compatibility .
How to begin with OpenAI through Google Cloud Marketplace
OpenAI models are purchasable on Google Cloud Marketplace and customers will handle the relevant usage charges on their existing Google Cloud accounts. Google says this spend may be eligible for Flexible Savings Plans and eligible cloud commitments.
This setup could reduce procurement work for organizations already using Google Cloud. It also allows development teams to compare model costs without having to establish separate commercial relationships with multiple providers.
Google hasn’t yet released full public pricing or independent performance benchmarks for the three models. Companies will need this information to understand exactly how much it will cost to run their applications.
Enterprise Data Protection and Controls
According to Google, OpenAI workloads will run on dedicated Google Cloud infrastructure, with customer requests routed through isolated gateways. The company says the responses, support information and prompts it generates are not used to train foundation models and stay within the bounds of the data it has access to.
Organizations have the ability to control access to models using authenticated identities, audit logging, and VPC Service Controls. These safeguards are designed to help IT teams monitor how AI is being used and limit access to sensitive data.
And Google has also revealed plans to add support for Confidential Compute and Zero Data Retention options. Both are expected features, not confirmed capabilities of the first preview.
Availability and Private Preview coming soon.
To start, GPT-Astra, GPT-6.1 Sol and GPT-6 Luna will be made available to a select group of enterprise customers via Google Cloud’s Model Garden. Organizations that are interested can request private preview access.
Developers will be able to access the models via Google’s APIs and integrate supported Codex CLI and desktop workflows with Google Cloud for inference, the company said. Google Antigravity and Gemini for business will be integrated with OpenAI models, though the company has not yet announced when it will be launched.
The addition provides Google Cloud customers with another way to run AI applications that is not solely tied to Google’s own models. Its practical value will depend on how well OpenAI’s models perform in real business workloads, as well as their competitive pricing and availability.