OpenAI models on Google Cloud: one invoice, keep routing portable
GPT-6 Astra, Sol and Luna hit Google Cloud Model Garden in private preview, billed on your GCP invoice. Here's how to use it without new lock-in.
OpenAI models are now on Google Cloud. On October 21, Google put GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna into Model Garden as a private preview for select enterprise customers, billed through Google Cloud Marketplace. Google says that makes it the only hyperscale cloud selling Gemini, Claude and GPT side by side. For a business that already pays Google for compute, the interesting part isn't the models. It's the invoice.
What actually happened
Google's announcement and OpenAI's companion post lay out the terms:
- Three models in private preview. Astra for heavy coding, computer use and research. Sol for near-Astra work at lower cost. Luna for fast, high-volume everyday tasks. Access starts with an interest form. OpenAI says general availability is planned in the coming weeks.
- Billing through Marketplace. Usage lands on your single Google Cloud invoice, draws down existing cloud commitments, and counts toward Flexible Savings Plans across model providers. No separate OpenAI contract.
- Codex now, more later. Developers can point local Codex CLI and Desktop workflows at Google's platform for inference today. Google Antigravity and Gemini for business support are listed as "coming soon."
- Data terms. Prompts and outputs stay inside your data perimeter and are not used for training. Zero Data Retention and Confidential Compute are both marked "coming soon," not live.
- No public per-token prices on Google's page.
Why OpenAI on Google Cloud matters for your business
Committed spend just got more flexible. If you signed a GCP commit and you're under it, GPT usage can now burn it down. That's real money for a shop that was paying OpenAI directly on a card while Google credits sat idle.
Don't trade one lock-in for another. Moving every call into Vertex because the bill is convenient makes Google your switching cost. Keep model selection behind your own routing layer: one interface, provider and model as config. Then Google, Azure, AWS or OpenAI direct becomes a price comparison, not a migration.
Wait for ZDR if you handle sensitive data. Customer records, payment details, health notes: the zero-retention option isn't live yet. Check the terms on the day you move, not the press release.
Price it before you switch. With no public rates, compare your actual monthly token bill against what Google quotes you, including any commit discount.
Key takeaways
- GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are in private preview in Google Cloud Model Garden as of October 21, 2026
- Usage bills through Google Cloud Marketplace and draws down existing commits and Flexible Savings Plans
- Codex CLI and Desktop can use Google's platform for inference now; Antigravity and Gemini for business come later
- Zero Data Retention and Confidential Compute are not live yet
- Keep provider choice in your own routing layer so the cheapest cloud wins each quarter
The model is a config line. The system around it should be yours. We build vendor-agnostic AI workflows with routing, cost tracking and fallbacks you own, so a new cloud deal is a switch, not a rewrite. See how we build, or send us the stack you're paying for twice.
Sources: Google Cloud blog, OpenAI.
- #openai
- #google-cloud
- #vertex-ai
- #model-routing
- #vendor-lock-in
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
Town raises $108M: AI assistants for teams need sharing rules
Town hit a $1B valuation with a $108M Series B to let AI assistants pass work between teammates. Write down what each agent can share before it starts talking.
Read itAI agent attack hit 2,250 Microsoft 365 tenants: audit your OAuth apps
Microsoft says Storm-0940 used the ARTEX AI agent framework to reach 2,250+ Microsoft 365 tenants via ConnectWise apps. Audit your OAuth app grants now.
Read it