OpenAI brings GPT-6 models and Codex to Google Cloud

Key takeaways

• OpenAI's frontier models, GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna, are rolling out in private preview in Google Cloud's Model Garden for select enterprise customers. General availability is planned "in the coming weeks."
• Customers will buy the models through Google Cloud Marketplace on a single Google Cloud invoice, and that usage can draw down existing Google Cloud spending commitments.
• Codex users can already route inference from the Codex CLI and desktop app through Google's platform. OpenAI models are also due "soon" in Google Antigravity and Google's Gemini agent offering.
• The deal follows Microsoft's switch to a non-exclusive licence in April 2026. OpenAI's models are now offered on Microsoft Azure, Amazon Web Services and Google Cloud.
• Microsoft remains OpenAI's primary cloud partner, and its April agreement says OpenAI products ship first on Azure unless Microsoft cannot and chooses not to support them.

What was announced

OpenAI and Google Cloud announced on Wednesday, October 21, that OpenAI's GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna models and its Codex coding agent will be sold through Google Cloud. The models are entering private preview in Model Garden for select enterprise customers.

Google's post names three OpenAI models in the first wave. GPT-6 Astra is aimed at demanding coding, computer use, professional and scientific work. GPT-6.1 Sol is described as offering "near-Astra intelligence at lower cost" for agentic coding and professional tasks. GPT-6 Luna is a faster, cheaper model for everyday high-volume use. Google says the three models are now rolling out in private preview to select enterprise customers and that more OpenAI models are "coming soon." Interested customers are directed to an interest form. According to The Next Web and Reuters, OpenAI expects general availability within weeks.

The models appear in Model Garden, Google Cloud's catalogue for discovering, testing and deploying models. Google says the catalogue now holds more than 200 first-party, third-party and open models. With OpenAI added, Google says it is the only hyperscale cloud provider offering Gemini, Anthropic's Claude and OpenAI's GPT models on one enterprise platform.

Codex, Antigravity and Gemini

The agreement also covers coding agents. Google says customers can connect their local Codex CLI and Codex desktop workflows to Google's Gemini platform so that inference runs on Google Cloud. According to OpenAI's announcement, as summarised by several outlets, OpenAI models will also arrive soon in Google Antigravity and the Gemini Agent. Antigravity is Google's agent-first development environment, which launched in November 2025. Google's own post refers to this as bringing the models to Antigravity and to Gemini for business. Neither company has given a date for these integrations.

Billing and commitments

The commercial terms may matter more than the technical ones for large buyers. Google says OpenAI models are bought directly through Google Cloud Marketplace and billed on the customer's existing Google Cloud invoice. Usage draws down existing cloud commitments and counts toward Flexible Savings Plans across model providers. Many large companies have already committed to multi-year spending with a cloud provider. If they can apply that money to OpenAI models, they can use the models without signing a separate vendor contract or going through a new procurement process. Reuters noted that cloud providers usually take a share of sales made through their marketplaces. Neither company has disclosed pricing or a revenue split for this deal.

How OpenAI got to all three clouds

Under its earlier agreements, Microsoft held exclusive rights to sell OpenAI's models as a cloud service, and Azure OpenAI was the only hyperscaler route for enterprises. That changed on April 27, 2026. Microsoft then said in a blog post that its licence to OpenAI intellectual property for models and products runs through 2032 but "will now be non-exclusive." The same post said that "OpenAI can now serve all its products to customers across any cloud provider."

The April terms also changed how money flows between the companies. Microsoft said it no longer pays a revenue share to OpenAI. OpenAI's revenue-share payments to Microsoft continue through 2030 at the same percentage, subject to a total cap that was not disclosed. Microsoft remains OpenAI's primary cloud partner. According to the post, OpenAI products ship first on Azure unless Microsoft cannot and chooses not to support the required capabilities.

Amazon moved first. On April 28, the day after the Microsoft announcement, AWS put OpenAI models, Codex on Amazon Bedrock, and Bedrock Managed Agents powered by OpenAI into limited preview. Amazon's own timeline shows several releases after that: GPT-5.6 Sol, Terra and Luna became generally available on Bedrock on July 9, and GPT-6 Astra became generally available there on September 8. Google Cloud is the third major provider to sign on.

Google and OpenAI already had a supplier relationship. In 2025, Reuters reported that OpenAI had agreed to use Google Cloud for additional computing capacity, even though the two companies compete in consumer AI assistants and search.

The pitch to enterprise and regulated customers

Both companies frame the deal around governance. In Google's description, prompts, context and outputs stay within the customer's designated data perimeter and are "never used to train foundation models." Google also says the OpenAI models run on dedicated Google Cloud infrastructure, that queries pass through isolated gateways, and that every model and agent interaction runs under an authenticated identity with VPC Service Controls and unified audit logging. Two features that matter to sensitive workloads are not yet available. Google lists zero data retention options and Confidential Compute as "coming soon."

Reuters reported that customers in regulated industries can use the models with Google Cloud's access controls, data governance tools and infrastructure to build agentic AI workflows. Banks, insurers, healthcare providers and public-sector agencies often have to approve each new data processor. For them, using a model inside a cloud they have already approved can be simpler than taking on a new direct supplier. Some banks are already reorganising around AI: Barclays plans to cut about 5,000 roles as it leans on AI. Customers will still need to confirm which controls apply during the private preview, and which regions and compliance certifications cover the OpenAI models. Google's post does not specify regions. For developers wiring such agents to internal systems, our tutorial on how to build an MCP server in Python shows one common way to expose company tools and data to a model.

Google Cloud chief executive Thomas Kurian said the company is focused on giving customers a broad selection of models, including OpenAI's, according to The Next Web. OpenAI president Greg Brockman called the partnership "an important step" in making the company's most capable models and Codex available to more customers.

What it means

For OpenAI, the deal widens distribution to customers whose budgets and data already sit inside Google Cloud. It also puts OpenAI in the same position as Anthropic, which already sells through more than one hyperscaler. Reuters, citing Anthropic's IPO filing, reported that 47% of Anthropic's 2025 revenue came through Amazon's and Google's cloud platforms. That figure shows how much enterprise AI spending now flows through cloud marketplaces. Wider distribution also matters for OpenAI's financing story; see our report on OpenAI nearing a $30 billion raise at a $1.25 trillion valuation. Older enterprise vendors are chasing the same budgets, as IBM's third-quarter results, with software up 8%, illustrate.

For Google, the move is a bet that being a neutral platform is worth more than keeping rival models out. It also brings risk for its own models. According to The Next Web, Moor Insights & Strategy analyst Mike Leone told Axios that Gemini "now has to win those customers with OpenAI sitting right next to it."

For Microsoft, Azure loses its status as the only hyperscaler that sells OpenAI's flagship models. It keeps first-ship rights, a licence through 2032 and revenue-share payments through 2030.

For enterprise buyers, choosing a model and choosing a cloud are becoming separate decisions. A company can now pick a cloud for its data, security and existing contracts, then choose among Gemini, Claude and GPT models inside it. Several details are still unknown: pricing on Google Cloud, the general availability date, which further OpenAI models will follow, and when the Antigravity and Gemini integrations, zero data retention and Confidential Compute will launch. Microsoft, Alphabet and Amazon report quarterly results next week, which may give the first signs of how the change is affecting their cloud businesses.

Related coverage

• TorchTPU and TorchNeuron explained: the PyTorch push past Nvidia CUDA
• How to Use Gemini 4 Argon Through the API: A Step-by-Step Guide
• How to Run a Local LLM with Ollama: Step-by-Step Guide
• OpenAI says its coding agents weakened internal AI monitoring

Sources

• Google Cloud Blog: Google Cloud expands model choice and flexibility, bringing OpenAI models to customers
• OpenAI: OpenAI models are coming to Google Cloud
• Reuters via The Star: OpenAI to sell its top models on Google Cloud in latest enterprise push
• The Next Web: OpenAI brings GPT-6 Astra and Codex to Google Cloud, a rival's platform
• Microsoft: The next phase of the Microsoft-OpenAI partnership
• AWS What's New: Amazon Bedrock offers OpenAI models, Codex and Managed Agents (limited preview)
• About Amazon: OpenAI models on Amazon Bedrock