Two of the biggest rivals in artificial intelligence are now partners. On October 21, 2026, OpenAI and Google Cloud announced that OpenAI's GPT-6 models and Codex workflows are coming to Google Cloud. With this deal, OpenAI's models are now offered through all three major U.S. cloud providers.
What Was Announced: OpenAI Models Arrive on Google Cloud
In what is easily the biggest OpenAI latest news story of the month, Google Cloud will offer OpenAI's frontier models in its Model Garden, the catalog where it already lists more than 200 models from Google, partners and open-source projects. Enterprise customers can buy access through Google Cloud Marketplace and have the usage count toward their existing Google Cloud spending commitments.
The deal also covers developer tools. Google says customers can connect local Codex CLI and Codex Desktop workflows to its platform for inference, starting on the day of the announcement. Further integrations with Gemini for business and Google Antigravity, Google's agentic coding environment, are described as coming later.
In one line: Teams that run on Google Cloud can now use OpenAI's models alongside Gemini, with one bill and one set of security controls.
The OpenAI Google Cloud Partnership 2026 at a Glance
- Announcement date (2026)
- Oct 21
- Announcement date (2026)
- GPT-6 family models announced
- 3
- GPT-6 family models announced
- Models in Google's Model Garden
- 200+
- Models in Google's Model Garden
- Current status (GA in weeks)
- Preview
- Current status (GA in weeks)
- CLI & Desktop workflows supported
- Codex
- CLI & Desktop workflows supported
- Big U.S. clouds now offer OpenAI
- 3 of 3
- Big U.S. clouds now offer OpenAI
Based on the October 21, 2026 announcement and press coverage. Rollout details may change.
Which OpenAI Models Are Coming to Google Cloud?
The launch lineup is OpenAI's current GPT-6 family. Together, these models give Google Cloud customers a range of price and performance options:
🚀 GPT-6 Astra
OpenAI's most capable model, built for demanding coding, computer use, professional work and scientific research. It is the first model being brought to Google Cloud.
☀️ GPT-6.1 Sol
Close to Astra-level intelligence at a lower cost. Reports put its price at roughly a fifth of Astra's.
🌙 GPT-6 Luna
Built for fast, high-volume everyday tasks. Reports put it at roughly 100 times cheaper than Astra.
Google Cloud now lists Gemini, Claude and GPT-6 in one catalog. For teams comparing Google Cloud AI models, that means they can test all three leading closed model families inside the same project, under the same IAM and logging setup.
The OpenAI Codex Coding Agent on Google Cloud
For many engineering teams, the Codex part of the deal matters most. The OpenAI Codex coding agent can read a repository, write and refactor code, run tests and propose changes. With this partnership, those workflows can use Google Cloud for inference:
- 1Codex CLI: Developers keep their terminal-based agent and send its model calls through Google Cloud.
- 2Codex Desktop: The desktop app can use the same Google Cloud route, so usage stays on one invoice.
- 3Governance: Spending, audit logs and access policies stay inside the company's existing Google Cloud setup.
This feature isn't exclusive to Google: AWS added a similar Codex setup on Amazon Bedrock in September 2026. Even so, it removes a common objection from platform teams that work only on Google Cloud.
Why OpenAI and Google Are Working Together Now
What OpenAI Gets
- • Access to enterprises that run only on Google Cloud
- • A way into large, pre-committed cloud budgets
- • Less reliance on a single distribution partner
What Google Cloud Gets
- • A model catalog that now includes GPT-6
- • Keeps customers who might otherwise move workloads to Azure or AWS
- • More usage running on its AI infrastructure
Google Cloud CEO Thomas Kurian said Google is "focused on giving customers a broad selection of models." The broader trend in AI cloud computing is that enterprises want to pick models freely, and cloud providers now compete on governance, billing and infrastructure rather than on having exclusive models. OpenAI says Microsoft remains its primary cloud partner, and Azure still gets new OpenAI launches first.
OpenAI on Azure vs AWS vs Google Cloud
| Platform | GPT-6 status | Notable for |
|---|---|---|
| Microsoft Foundry (Azure) | GA since Sept 2026 | Primary partner, first to get new launches, published list prices |
| Amazon Bedrock (AWS) | GA since Sept 2026 | Codex support, zero data retention available on request |
| Google Cloud Model Garden | Private preview (Oct 2026) | Only one that also offers Gemini, plus commitment drawdown |
| OpenAI API (direct) | Available | Newest features first, OpenAI's own pricing and tooling |
Benefits for Enterprise AI Solutions
For large organizations, the main benefit is simpler procurement. Before this deal, using OpenAI models on Google Cloud usually meant a separate vendor contract, a separate budget and a security review of a new data flow. Now, according to Google:
- One invoice: OpenAI usage appears on the Google Cloud bill.
- Commitment drawdown: Spending counts against existing Google Cloud commitments.
- Data perimeter: Prompts, context and outputs stay inside the customer's environment and are not used for model training.
- Security controls: Isolated gateways, VPC Service Controls and unified audit logging.
- Coming soon: Zero data retention and Confidential Compute based on trusted execution environments.
For regulated industries such as finance, healthcare and the public sector, these controls are often what decides whether an enterprise AI solution moves past the pilot stage.
What It Means for AI Developers
More Choice, Less Lock-In
You can send each task to Gemini, Claude or GPT-6 without leaving your cloud project. That makes model routing and side-by-side evaluations much easier.
Simpler OpenAI API Integration
If your app already runs on Google Cloud, OpenAI API integration can use the same IAM, networking and logging you already have, so you don't need to manage separate keys for each vendor.
Agentic Coding at Scale
Running Codex through Google Cloud makes it easier to roll AI-powered software development out to whole teams with central budgets and controls.
Cost Tiers for Every Workload
Use Astra for hard reasoning, Sol for most production traffic and Luna for high-volume, latency-sensitive tasks.
For anyone building AI tools for developers, the takeaway is that models are increasingly available everywhere. What sets products apart is how well they combine models, evaluations, data and workflow, not which provider they are tied to.
Current Limitations and Open Questions
Preview only
Access currently requires an interest form, and pre-GA services usually come without an SLA.
No public pricing yet
Google hasn't published per-token rates. AWS reportedly charges a regional premium over OpenAI's direct pricing, so compare before you switch.
Unpublished details
Supported regions, model identifiers and quotas haven't been fully documented.
Drawdown is not a discount
Third-party marketplace products typically don't get extra savings-plan discounts, so the benefit is using committed budget, not a lower per-token rate.
Data retention options pending
Zero data retention and Confidential Compute are listed as coming soon.
Developer Action Plan: What to Do Now
| If you are… | Recommended next step |
|---|---|
| A Google Cloud team exploring GPT-6 | Submit the interest form and run evaluations against your current models during the preview. |
| Planning production workloads | Wait for GA, published pricing, regions and an SLA. |
| In a regulated industry | Get the zero data retention scope and timing from Google in writing before you commit. |
| Rolling out Codex to engineers | Pilot Codex CLI through Google Cloud with one team, measure spend, then expand. |
| Already on Azure or AWS | Compare per-token costs before moving workloads just to use committed budget. |
💡 Key Takeaway
The OpenAI Google Cloud partnership 2026 is less about two rivals becoming friends and more about enterprise demand. Customers want the best model for each task, inside the cloud they already trust. For developers, that means more choice, simpler governance and faster adoption of AI-powered software development.
Frequently Asked Questions
What is the OpenAI and Google Cloud partnership?
+
Announced on October 21, 2026, the partnership brings OpenAI's frontier models to Google Cloud's Model Garden, so Google Cloud customers can use them through Google Cloud Marketplace. It also lets developers connect Codex CLI and Codex Desktop workflows to Google's platform for inference.
Which OpenAI models are available on Google Cloud?
+
The announced lineup is GPT-6 Astra, OpenAI's flagship model, plus the lower-cost GPT-6.1 Sol and GPT-6 Luna. GPT-6 Astra is the first model being brought to the platform. Check Google Cloud's Model Garden for the exact models and identifiers available in your region.
Are OpenAI models generally available on Google Cloud yet?
+
Not yet. At launch, the models were rolling out in private preview to selected enterprise customers through an interest form. OpenAI says general availability is planned in the coming weeks, but no exact date has been published.
Can I use my Google Cloud credits or commitments to pay for OpenAI models?
+
Yes. Purchases go through Google Cloud Marketplace, appear on your Google Cloud invoice and can draw down existing cloud spending commitments. Third-party products do not usually get extra savings-plan discounts, so the main benefit is using budget you have already committed, not a lower per-token price.
Does this mean OpenAI is leaving Microsoft Azure?
+
No. OpenAI says Microsoft remains its primary cloud partner, and Azure says OpenAI continues to launch first on Azure. Google Cloud is an additional distribution channel, following Microsoft Foundry and Amazon Bedrock.
How does the OpenAI Codex coding agent work with Google Cloud?
+
Developers can point their local Codex CLI and Codex Desktop workflows at Google's platform for model inference. Usage can then be billed through Google Cloud and kept inside the company's Google Cloud governance controls.
Is my data used to train models when I use OpenAI on Google Cloud?
+
According to Google, prompts, context and outputs stay inside the customer's data perimeter and are not used to train foundation models. Zero data retention and Confidential Compute options are described as coming soon, not available at launch.
Related reading: Anthropic IPO 2026: Latest News, AI Market Growth and Investor Outlook
Sources
- OpenAI Developer Community: OpenAI models are coming to Google Cloud
- SiliconANGLE: OpenAI to make its most advanced AI models available in Google Cloud
- TechNode Global: Google Cloud to offer OpenAI models through Model Garden
- Techstrong.ai: OpenAI partners with Google Cloud to extend enterprise distribution
- Digital Applied: OpenAI models come to Google Cloud, what is in preview