OpenAI’s models are coming to Google Cloud and Antigravity. What changes for Codex

What was announced

On October 21, 2026, OpenAI and Google Cloud said they are “partnering to bring OpenAI’s frontier models and Codex agentic workflows into Google Cloud’s Model Garden.” Google’s announcement names three models:

• GPT-6 Astra, for “demanding work across coding, computer use, professional tasks, and scientific research.”
• GPT-6.1 Sol, “near-Astra intelligence at lower cost.”
• GPT-6 Luna, “a fast, cost-efficient model for everyday work at scale.”

Companies buy them through Google Cloud Marketplace, so the usage appears on their Google Cloud invoice and draws down the cloud commitments they already have. OpenAI’s announcement adds that Microsoft “remains OpenAI’s primary cloud partner.”

What’s available, and when

What	Status on October 22
GPT-6 Astra, 6.1 Sol and 6 Luna in Model Garden	Private preview, select enterprise customers
General availability on Google Cloud	“In the coming weeks,” says OpenAI
Codex CLI and desktop on Google’s inference	Google says from October 21; no setup docs yet
OpenAI models in Google Antigravity	“Soon,” no date
OpenAI models in Gemini for business	“Soon,” no date
Zero data retention, Confidential Compute	“Coming soon”

To get in now, you register interest through Google’s form or OpenAI’s. Neither company published Google Cloud prices. OpenAI’s own API prices, per million tokens, are $10 in and $50 out for Astra, $2 and $10 for Sol, and $0.10 and $0.50 for Luna, each with a context window of 1.05 million tokens.

What it means for Codex users

If you use Codex on a ChatGPT plan, nothing changes. Neither announcement mentions ChatGPT plans, their limits or their prices.

If your company runs on Google Cloud, Codex may soon bill there instead. Google says customers can “connect their local Codex CLI and Desktop workflows directly to our Gemini platform for inference.” Codex’s documentation had no Google Cloud page by October 22, but it has one for Amazon Bedrock, and that shows the likely shape:

• Codex points at the cloud with model_provider in ~/.codex/config.toml, and the CLI, the desktop app and the editor extension all read it.
• It signs in with the cloud’s credentials, not ChatGPT, and its model requests go to the cloud rather than OpenAI’s servers.
• Features that live in OpenAI’s cloud don’t come along. On Bedrock that includes Codex cloud tasks, Fast mode, image generation and plugins that need a ChatGPT sign-in.

Expect the Google version to work the same way, and check its documentation when it appears: this part is our reading of the Bedrock precedent, not something either company has said.

What it means for Antigravity users

Not much yet. As of October 22, Antigravity’s models page lists no OpenAI frontier model. The one OpenAI model it has, the open GPT-OSS-120b, is removed from every plan on November 2, along with Claude Sonnet 4.6 and Opus 4.6 (see our memo on that change). Neither company has said which Antigravity plans will get GPT when it arrives.

Using GPT and Gemini together today

You don’t have to wait for one app to carry both labs. Enjoy runs Codex on your ChatGPT plan and Antigravity on your Google account side by side, with Claude Code and the other labs’ agents beside them. Start a job with Gemini, switch the conversation to Codex when GPT suits it better, and the work comes with you. Enjoy is free; each agent runs on the plan you already have.

We’ll update this memo when Google publishes the Codex setup or GPT arrives in Antigravity.