GPT-6 Astra, Sol and Luna Land on Google Cloud in Private Preview, Antigravity Next

News Summary

OpenAI's frontier models are now arriving on Google Cloud, a notable shift for the AI industry's infrastructure landscape. On October 21, 2026 (Pacific Time), OpenAI and Google announced that Google Cloud customers will be able to buy OpenAI models through Google Cloud's Model Garden, with billing handled through Google Cloud Marketplace. A Google Cloud blog post dated October 22, 2026, then detailed three models in private preview, while Google Antigravity and Gemini for business are listed as "coming soon." This report cross-checks the Google Cloud announcement against coverage from 9to5Google, Slashdot, and other outlets. The Ars Technica article that prompted this report could not be retrieved, so its contents are not used here.

What Was Announced

According to Google Cloud's own blog post, written by Michael Gerstenhaber, VP of Product Management for Gemini at Google Cloud, three OpenAI models are rolling out in private preview to select enterprise customers: GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna. Google says more models are coming soon. Customers who want access can register interest through a form on Google Cloud's website. The models can be used through Google Cloud APIs starting now, and customers can also connect local Codex CLI and Codex Desktop workflows to Google's Gemini platform for inference.

Note that naming differs slightly across coverage. One outlet referred to the smallest model as "GPT-6.1 Luna," while Google's blog uses "GPT-6 Luna." The 9to5Google report said the exact models had not been confirmed at the time of OpenAI's announcement. This report follows Google's primary-source naming.

The Three Models

Google describes GPT-6 Astra as the model for demanding work across coding, computer use, professional tasks, and scientific research. GPT-6.1 Sol is positioned as delivering near-Astra intelligence at lower cost, aimed at agentic coding, computer use, and professional work. GPT-6 Luna is described as a fast, cost-efficient model for everyday work at scale. In practice, this tiering lets teams match each workload to a model by performance, latency, and cost.

Billing and Procurement

OpenAI models are transacted directly through Google Cloud Marketplace. Usage appears on a single Google Cloud invoice, draws down existing cloud spending commitments, and counts toward Flexible Savings Plans across model providers. Customers do not need separate vendor contracts. For large organizations, that procurement simplicity may matter as much as the models themselves.

Security and Governance

Google states that prompts, context, and outputs stay within the customer's designated data perimeter and are never used to train foundation models. Zero Data Retention options and Confidential Compute are both listed as coming soon. The OpenAI models run on dedicated Google Cloud infrastructure behind hardened, isolated gateways. Each agent and model interaction runs under an authenticated identity, with VPC Service Controls protection and unified audit logging. Box's VP of AI Products, Yashodha Bhavnani, said the arrangement will let customers use OpenAI's frontier models alongside Google Cloud's security and compliance capabilities.

Antigravity and Gemini for Business

Google Antigravity, Google's agentic coding tool, will gain OpenAI models "soon," according to the announcement, with no specific date. 9to5Google reported that OpenAI models are also coming to Gemini Agent, and Google's blog lists Gemini for business as a coming-soon destination. 9to5Google noted that Antigravity currently offers Anthropic models: Claude Opus and Sonnet 4.6 for free, Plus, and trial accounts until November 2, and Opus and Sonnet 5.5 for paid Pro and higher tiers. Cloud availability beyond private preview was described by OpenAI as arriving "in the coming weeks."

Context: A Multi-Cloud Era for OpenAI

Slashdot's summary of CNBC reporting says Microsoft previously held exclusive licensing rights to OpenAI's intellectual property, and that exclusivity ended earlier in 2026. Amazon then announced plans to host OpenAI models on AWS. With Google Cloud now added, OpenAI's models are available across all three major U.S. cloud platforms. Coverage also reported OpenAI saying Microsoft remains its primary cloud partner. OpenAI said Google Cloud's longstanding investment in machine learning has made it a destination for AI workloads.

Google Cloud also highlights that it is the only hyperscale provider offering Gemini, Claude, and GPT models, alongside more than 200 first-party, third-party, and open models in Model Garden.

Why It Matters for Developers

For engineering teams, the practical takeaway is flexibility. Model Garden lets developers swap models with less architectural lock-in, and run agentic workloads across different providers under one security and billing setup. Caveats remain: access is currently private preview, general availability timing is vague, and Antigravity support has no date. Readers should verify current availability on official Google Cloud and OpenAI pages before planning deployments.

Sources Consulted

Google Cloud Blog (October 22, 2026); 9to5Google (October 21, 2026, 10:30 a.m. Pacific Time); Slashdot (October 21, 2026), summarizing CNBC; additional headlines from Axios, PYMNTS, and Storyboard18. Axios and CNBC pages were not directly accessible for this report.