OpenAI brings frontier models and Codex to Google Cloud's Model Garden

OpenAI frontier models are getting a second major cloud storefront. In an announcement dated October 21, 2026, OpenAI confirmed a partnership with Google Cloud that brings its models and Codex agentic workflows into Gemini's Model Garden, with purchases transacted through Google Cloud Marketplace and general availability planned in the coming weeks. Google's pitch, per Ars Technica, is simple: Google Cloud becomes the only major provider offering all three leading closed model families, Gemini, Claude, and OpenAI, on one bill.

OpenAI's models enter the Model Garden

The official post describes "native access" to OpenAI's frontier models and Codex inside the environment where Google Cloud teams already build, aimed at enterprise software development, agentic workflows, and regulated industries that need centralized access controls and data governance. OpenAI does not name specific models in the announcement. Ars Technica reports that the full suite of current GPT 6 and 6.1 models, Astra, Sol, and Luna, will launch on Google Cloud, joining the garden's existing 200-plus models.

Marketplace billing puts GCP commitments behind GPT

The commercial mechanics may matter more to buyers than the model names. Customers will purchase OpenAI models through Model Garden via Google Cloud Marketplace, and enterprises with Google Cloud spend commitments can draw those commitments down on OpenAI usage. For teams holding large committed-use agreements, that lowers the friction of testing or running GPT workloads without a separate OpenAI invoice. The OpenAI post also notes Codex access: Ars Technica reports that an early-access request now enables API usage even when you drive it from Codex CLI or Codex Desktop, with inference running on Google's infrastructure.

Antigravity and the Gemini Agent come next

Availability is staged, not fully live. OpenAI says its models will "soon" appear in Google Antigravity, the agent-powered coding client, and in the Gemini Agent, offering model choice per task. Ars Technica adds that Google has not yet specified which OpenAI models will show up in Antigravity, which currently supports Gemini and Claude. Treat every "coming soon" surface as unconfirmed until it appears in product.

A multi-cloud milestone, with two caveats

First, OpenAI is explicit that Microsoft remains its primary cloud partner. This is expanded distribution, not a divorce. Second, the launch joins a pattern rather than breaking one: Claude already reaches users through Amazon Bedrock and Google Vertex AI, and OpenAI's own models recently landed at Vercel AI Gateway. What makes this one substantial is the enterprise plumbing, Marketplace billing, commitment drawdown, Codex integration, and Google's first-party coding-agent surfaces, rather than mere API availability.

What to verify before you act

If you are considering routing GPT workloads through Google Cloud, check four things at GA time: the exact model IDs listed in Model Garden, pricing and rate limits versus the OpenAI API, whether your GCP commitment drawdown applies to Marketplace model purchases under your contract terms, and which Antigravity and Gemini Agent surfaces actually ship. Until general availability lands in the coming weeks, this is an announced rollout with an access-request page, not a switch you can flip today. For a broader map of agent tooling across providers, see LinkLoot's AI agent tools guide.

Source check

Claims about the partnership, billing, quotes from Greg Brockman and Thomas Kurian, and the "coming weeks" GA window come from OpenAI's official announcement, which LinkLoot read in full via an isolated public browser. Model naming (GPT 6 and 6.1 Astra, Sol, Luna), private-preview status, and Antigravity details come from Ars Technica's independent reporting. OpenAI's post itself does not name the specific models.