Google Cloud is adding the OpenAI models GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna to its model garden. The models are available from today in private preview to a select group of business customers. Usage runs through Google Cloud Marketplace and appears on the regular Google Cloud bill.
Michael Gerstenhaber, VP Product Management for Gemini at Google Cloud, announced the move in a blog post. According to Google Cloud, this step makes it the only hyperscaler to offer access to models from all three major providers of closed-source LLMs. These sit on a single AI platform alongside more than 200 proprietary, third-party and open models. Gerstenhaber argues that enterprise AI applications call for selecting the right model for each workload, rather than using a single model for every task.
Three models for different workloads
According to Google Cloud, GPT-6 Astra is intended for demanding tasks in programming, computer use, professional work and scientific research. GPT-6.1 Sol is said to deliver performance close to that of Astra at a lower cost. That model is aimed at agentic coding, computer use and professional work. The company describes GPT-6 Luna as a fast, cheaper model for everyday tasks at scale.
Google Cloud says customers can switch between models for each use case based on performance, latency and cost. According to the company, this requires no changes to the architecture, and customers do not need to sign separate contracts with model providers.
Billing via Marketplace
The OpenAI models are purchased through Google Cloud Marketplace. As a result, usage appears on the same Google Cloud bill as other cloud services and is drawn down against existing cloud commitments. It also counts towards the Flexible Savings Plans, which Google Cloud applies across multiple model providers.
Security and data governance
Google Cloud states that customers’ prompts, context and output remain within the data perimeter designated by the customer. According to the company, this data is not used to train foundation models. Zero Data Retention options will follow later.
According to Google Cloud, the OpenAI models run on dedicated, segregated infrastructure within Google Cloud. Support for Confidential Compute will become available later. All requests pass through isolated gateways, so that data traffic stays within the Google Cloud environment. The company says this means external model providers cannot gain unauthorised access to the customer’s cloud environment.
Every interaction with an agent or model takes place under an authenticated identity. According to Google Cloud, administrators gain visibility into the entire chain, protection through VPC Service Controls and centralised audit logging across the whole organisation.
Box to deploy the models
Document platform Box plans to offer the OpenAI models to its customers via Google Cloud. Yashodha Bhavnani, VP of AI Products at Box, says this will allow the company to combine OpenAI’s models with Google Cloud’s security and compliance features. Bhavnani expects customers to be able to get more out of their business content as a result.
Availability
The three models are now rolling out in private preview to selected business customers. Google Cloud says more models will follow. Customers can use the models via Google Cloud’s APIs. They can also connect their local Codex CLI and Codex Desktop workflows to the Gemini platform for inference. Those interested can register via a form.
In due course, the OpenAI models will also become available in Gemini for business and in the Google Antigravity development environment.
This article was originally published in Dutch on MSP Business. Read the original →
