Google Cloud is adding OpenAI’s GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna models to its Model Garden, expanding the AI options available to enterprise developers. Announced on October 21, 2026, the partnership allows eligible businesses to access OpenAI models through Google’s cloud infrastructure and APIs. The models are initially rolling out in private preview to select enterprise customers, with general availability planned for the coming weeks.
The partnership is designed to help organizations choose AI models based on their performance, cost, and speed requirements instead of relying on a single provider. Businesses will be able to purchase the models through Google Cloud Marketplace, with eligible spending applied to existing Google Cloud commitments. The announcement also expands the distribution of OpenAI’s models beyond its own platform and existing cloud channels, giving enterprises another option for deploying advanced AI applications.
Key Takeaways
- Three OpenAI models: Google Cloud is adding GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna to Model Garden.
- Private preview: Access is initially limited to selected enterprise customers.
- Different performance tiers: Astra targets demanding work, Sol balances capability and cost, and Luna focuses on fast, high-volume tasks.
- Unified billing: Purchases go through Google Cloud Marketplace and can draw down eligible existing cloud commitments.
- Enterprise security: Google Cloud provides access controls, governance, and audit capabilities around model usage.
- Codex integration: Developers can connect Codex CLI and desktop workflows to Google Cloud for model inference.
- General availability pending: Both companies say broader availability is planned, but a definitive public launch date has not been announced.
Which OpenAI Models Are Coming to Google Cloud?
The three models are designed to serve different enterprise workloads. This gives developers the flexibility to choose a model according to the complexity of a task, the required response time, and the cost of running it.
| Model | Main focus | Suitable workloads |
|---|---|---|
| GPT-6 Astra | Highest capability for demanding work | Complex coding, computer use, professional tasks, and scientific research |
| GPT-6.1 Sol | Near-Astra performance at a lower cost | Agentic coding, computer-use workflows, and business productivity |
| GPT-6 Luna | Speed and cost efficiency | Routine tasks, high-volume applications, and scalable AI features |
GPT-6 Astra
GPT-6 Astra is OpenAI’s flagship model in this lineup. Google describes it as suitable for demanding professional workloads, including complex software development, computer use, and scientific research.
For enterprises, this could mean using the model to support sophisticated AI agents, analyze complicated information, or handle multi-step coding tasks. Astra is intended for applications where accuracy and capability matter more than minimizing inference costs.
GPT-6.1 Sol
GPT-6.1 Sol is designed to deliver performance approaching Astra’s level while reducing costs. This makes it a potential option for organizations that need capable AI agents but want to manage their spending as usage grows.
Businesses could use Sol for coding assistance, computer-based workflows, and other professional tasks that require more reasoning than a lightweight model can reliably provide.
GPT-6 Luna
GPT-6 Luna prioritizes efficiency. It is designed for everyday work at scale, where response speed and cost per task can be as important as raw capability.
Potential applications include routine content processing, customer-support workflows, and other repetitive tasks. The appropriate model will depend on the complexity of the application and the quality requirements involved.
Why Google Cloud Is Adding OpenAI Models
The announcement reflects a shift in enterprise AI adoption. Businesses increasingly want access to multiple models because no single system is necessarily the best choice for every workload.
Google Cloud already offers Google’s Gemini models alongside a broader collection of first-party, third-party, and open models through Model Garden. Adding OpenAI’s newest models gives customers more options within the same cloud environment.
This approach could be particularly useful for companies that have already invested in Google’s infrastructure. They can evaluate OpenAI models without necessarily rebuilding their applications around a separate cloud platform.
It also strengthens Google’s position as a provider of infrastructure for AI applications, including those powered by competing AI developers. Rather than requiring customers to choose one model family for every task, Google is positioning its platform as a place where organizations can select models based on their needs.
Google says its platform offers more than 200 models across different providers and categories. Its October announcement described the OpenAI integration as an expansion of that model choice.
What the Partnership Means for Developers
The integration is intended to make it easier for developers to build applications with OpenAI models while working within Google Cloud.
According to the companies’ announcements, eligible customers can use the models through Google Cloud APIs. Developers can also connect local Codex CLI and desktop workflows to Google’s platform for inference, allowing supported coding workflows to use models through the Google Cloud environment.
This may reduce friction for teams that already manage their applications, identity systems, and cloud spending through Google. Instead of building separate infrastructure for every model provider, developers can work within a familiar environment.
The companies have also announced plans to bring OpenAI models to additional Google products, including Google Antigravity and Gemini-related enterprise experiences. Those integrations are future plans, not features that should be assumed to be available to every user at launch.
The immediate opportunity is for enterprise developers to evaluate the models, test them against existing workflows, and determine whether their capabilities justify the cost.
Billing and Enterprise Cloud Spending
A major commercial benefit is the ability to purchase OpenAI models through Google Cloud Marketplace.
For eligible businesses, usage can appear on a consolidated Google Cloud invoice. Purchases can also draw down existing Google Cloud spending commitments and count toward applicable Flexible Savings Plans, according to Google’s announcement.
This could simplify procurement for large organizations that already have negotiated cloud agreements. Companies may be able to evaluate OpenAI models without creating an entirely separate purchasing process, although access still depends on the preview program and the applicable commercial terms.
Neither company had published a definitive Google Cloud price list for the three models at the time of the announcement. Customers should therefore distinguish OpenAI’s direct API prices from the pricing that may apply when models are purchased through Google Cloud Marketplace.
Actual costs will depend on the commercial terms, model usage, and any applicable billing arrangements.
Security, Data Governance and Compliance
Enterprise adoption of AI depends not only on model quality but also on how sensitive information is handled.
Google says OpenAI models on its platform will operate with enterprise governance features, including authenticated identities, centralized audit logging, access controls, and support for Virtual Private Cloud Service Controls.
Google also says customer prompts, context, and outputs remain within the customer’s designated data perimeter and are not used to train foundation models. Zero Data Retention options and Confidential Compute capabilities were described as forthcoming enhancements rather than universally available features at launch.
These controls are intended to help businesses use advanced models while maintaining visibility into how employees and AI agents access enterprise resources. However, organizations should review the specific terms and technical documentation for their chosen service before processing regulated or highly sensitive information.
Security controls also do not eliminate every risk. Companies still need to manage permissions, application design, data access, and the actions AI agents are allowed to perform.
How This Changes the Cloud AI Market
The agreement illustrates the increasingly competitive and collaborative nature of the cloud AI market. Google and OpenAI compete in AI products, but their partnership allows OpenAI’s models to be distributed through Google’s enterprise infrastructure.
For OpenAI, Google Cloud offers another route to enterprise customers. For Google, the addition makes Model Garden more attractive to organizations that want model flexibility without abandoning their existing cloud environment.
The deal also reflects the growing importance of AI distribution. Building a powerful model is only part of the business challenge; providers must also make their systems accessible through developer tools, enterprise purchasing channels, and production infrastructure.
Microsoft remains an important OpenAI cloud partner, while OpenAI has continued expanding its distribution across other platforms. The Google Cloud agreement adds another option for customers deciding where and how to run AI workloads.
Availability and What Businesses Should Know
The three models are initially rolling out in private preview to select Google Cloud enterprise customers. Businesses interested in access can consult the Google Cloud announcement and use the company’s interest form.
General availability is planned for the coming weeks, according to OpenAI and Google Cloud. Neither company had announced a definitive public launch date in the cited releases.
Organizations evaluating the models should test them against representative workloads rather than choosing solely by headline performance. Coding quality, latency, reliability, cost, and integration requirements can vary substantially across applications.
The Bigger Picture
Google Cloud’s decision to host OpenAI’s frontier models strengthens the idea that enterprise AI platforms will increasingly offer models from competing developers side by side. Businesses want flexibility to choose the right balance of capability, price, speed, and governance for each task. Bringing the models into an existing cloud environment may make experimentation and deployment easier for organizations that already rely on Google infrastructure.
The long-term value of the partnership will depend on availability, pricing, performance in real-world workloads, and how well the models fit into enterprise development workflows. For OpenAI, the agreement expands distribution; for Google, it adds another reason for customers to use Model Garden. For businesses, the central benefit is greater choice without necessarily having to change their entire cloud strategy.
FAQs
1. Which OpenAI models is Google Cloud adding to Model Garden?
Google Cloud is adding GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna. The models target demanding reasoning and coding, cost-efficient professional work, and fast, high-volume tasks, respectively.
2. Are the models available to everyone?
Not yet. The models are rolling out in private preview to select enterprise customers. General availability is planned for the coming weeks, but a specific public launch date has not been announced.
3. How will customers pay for the models?
Purchases will go through Google Cloud Marketplace. Eligible customers can consolidate billing and use applicable existing Google Cloud commitments, subject to the relevant commercial terms.
4. Can developers use Codex with Google Cloud?
The companies say developers can connect their local Codex CLI and desktop workflows to Google Cloud for inference. Access and setup requirements depend on the supported workflow and preview availability.
5. Will Google Cloud use customer prompts to train AI models?
Google says customer prompts, context, and outputs remain within the designated data perimeter and are not used to train foundation models. Businesses should review the service’s current documentation and contractual terms for their particular deployment.
Looking Ahead
The addition of GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna gives Google Cloud customers more options for building AI applications and agentic workflows. If access expands as planned, enterprises will be able to compare OpenAI’s models with Gemini and other offerings within a broader cloud-based development environment. That could make it easier to choose different models for different tasks while maintaining a consistent approach to billing and governance.
The next milestones are broader availability, clear pricing, and evidence of how the models perform in production. Enterprises will need to measure whether the added capabilities translate into better results, lower costs, or faster workflows. The partnership is an important step in making frontier AI more accessible through cloud platforms, but its practical impact will depend on how well the integration meets real business needs.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
