OpenAI on Google Cloud: GPT-6 Astra, Sol, Luna in Preview

OpenAI’s models are coming to Google Cloud. The two companies announced on 21 October that OpenAI’s frontier models and Codex will be offered in Model Garden, Google’s catalogue of first-party, partner and open models. Google’s post names three: GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna, now rolling out in private preview to selected enterprise customers, with more models to follow. OpenAI says general availability is planned in the coming weeks. With them, Google says it becomes the only hyperscale cloud offering Gemini, Claude and GPT on one platform.

The commercial detail is the part enterprise buyers will notice. The models are bought through Google Cloud Marketplace and appear on the same Google Cloud invoice, and companies with spend commitments can draw them down on OpenAI usage, which also counts toward Google’s Flexible Savings Plans across model providers. Neither company published prices. Developers can call the models through the APIs, and Google says teams can point their local Codex CLI and Codex desktop app at the Gemini platform for inference. OpenAI models are also due later in Google Antigravity and in Gemini for business, including the Gemini Agent.

Google’s pitch leans on governance. It says prompts, context and outputs stay inside the customer’s own perimeter and are never used to train foundation models, that the OpenAI models run on dedicated Google Cloud infrastructure, and that every agent and model call runs under an authenticated identity with VPC Service Controls and unified audit logging. Zero data retention and Confidential Compute are both marked as coming soon. Box, an early customer, says the arrangement lets it offer OpenAI models inside Google Cloud’s security and compliance setup. Interested companies can apply through an interest form.

The third cloud this year

Google is the third of the big clouds to resell OpenAI’s models this year, after Amazon Bedrock in May and Oracle in June, and the deals follow one template: OpenAI models where the customer already has a contract, paid from a budget already committed. OpenAI’s post ends with a single line, “Microsoft remains OpenAI’s primary cloud partner”, which is about as much as either side said on that point. Greg Brockman, OpenAI’s president, framed it as bringing Codex “where customers are already building”; Thomas Kurian, Google Cloud’s chief executive, as letting customers pick the right model for each job.

The odder fit is on Google’s side. Antigravity is Google’s own coding tool, the one Google opened to Claude Opus 5 for its engineers in September, and it will now carry the models of Gemini’s most direct rival as well. For Google Cloud, a sale of GPT tokens is still a sale on its own invoice; for OpenAI, every cloud added is one less reason for a large customer to say no. The open questions are the ones the announcement skipped: the price, the regions, and the date the preview opens to everyone.