Google Cloud is adding OpenAI’s GPT-Astra, GPT 6.1 Sol and GPT 6.0 Luna to its model garden, bringing the models into its existing enterprise AI platform. Access is rolling out in private preview to select enterprise customers, with model usage managed through Google Cloud’s billing, security and governance systems.
The addition expands a catalogue of more than 200 first-party, third-party and open models. With their inclusion, Google Cloud claims it is now the only hyperscale cloud provider offering access to all three leading closed-source large language models.
Different models for different workloads
The three OpenAI models serve different computing requirements. GPT-Astra is designed for demanding tasks involving coding, computer use, professional work and scientific research. Intended to deliver intelligence approaching GPT-Astra at a lower cost, GPT 6.1 Sol covers similar tasks such as agentic coding and computer use, while GPT 6.0 Luna is positioned as a fast, cost-efficient option for everyday workloads at scale.
Through model garden, organisations can select and switch between models according to their workload requirements. Google Cloud says this flexibility will allow developers to choose different models based on performance, latency and cost efficiency without architectural lock-in or separate contracts with individual model providers.
OpenAI model usage will be transacted directly through Google Cloud Marketplace, with charges appearing on a single Google Cloud invoice. Usage will also draw down existing cloud commitments and count towards Flexible Savings Plans (FSP) across model providers.
Security controls and data handling
Google Cloud is applying its enterprise security and governance controls to OpenAI workloads. Customer prompts, context and outputs will remain within designated data perimeters and will not be used to train foundation models, according to the company.
OpenAI models will run on dedicated Google Cloud infrastructure, with requests routed through hardened, isolated gateways. These arrangements are intended to keep data transfers within Google Cloud’s secure environment and prevent third-party model providers from obtaining unauthorised access to customers’ cloud systems.
The platform also provides authenticated identities for model and AI agent interactions, allowing IT administrators to monitor activity through unified audit logs. VPC Service Controls protection and centralised administrative visibility are included as part of the governance arrangements.
Two additional protections are also planned. Google Cloud intends to introduce Zero Data Retention (ZDR) options and Confidential Compute support, both of which are described as coming soon.
Private preview and developer integrations
From 22 October, customers can build with the models through Google Cloud APIs and connect local Codex CLI and Desktop workflows directly to the Gemini platform for inference, the company says. Google Cloud also indicates that further availability is planned following the initial private preview.
Greg Brockman, President and co-founder of OpenAI, described the partnership as an opportunity to make the company’s models and Codex available where customers are already developing software and AI agents. Thomas Kurian, Chief Executive Officer of Google Cloud, said the expanded model selection would allow customers to choose models according to individual tasks, with the aim of improving quality and reducing costs.
Google Cloud also plans to integrate OpenAI models into Gemini for business and Google Antigravity. Both integrations are described as forthcoming, extending model choice to additional enterprise applications and developer tools.
