OpenAI brings its models and Codex to Google Cloud
GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna entered private preview in Google Cloud's Model Garden, putting OpenAI's models on all three of the largest US clouds.
What happened
On 21 October OpenAI and Google Cloud announced that GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are rolling out in private preview in Model Garden for select enterprise customers, with general availability planned "in the coming weeks". Codex can already connect to Google's platform, and the models are coming to Google Antigravity and the Gemini Agent.
What changes for customers
Customers buy through Google Cloud Marketplace, can count usage against existing Google Cloud commitments and get one invoice. Google says prompts and outputs stay within the customer's data perimeter and are never used for training. The models run on dedicated infrastructure. Pricing was not disclosed.
Rivals on one shelf
Google now says it is the only hyperscaler offering Gemini, Claude and GPT together. Microsoft remains OpenAI's primary cloud partner; AWS began hosting OpenAI models in April. Anthropic's IPO filing says 47% of its 2025 revenue came through Amazon's and Google's clouds. Analyst Mike Leone: "Gemini now has to win those customers with OpenAI sitting right next to it."
Why it matters
Model choice is decoupling from cloud choice. Google Cloud growth accelerated to about 82% in Q2, against 37% at Amazon and 42% at Microsoft; all three report next week.
Which model does what
The three models serve different needs. GPT-6 Astra, released in early September, is OpenAI's most capable model for long-context, complex reasoning work. GPT-6.1 Sol is the balanced general-purpose option, and GPT-6 Luna is the lighter model for high-volume tasks where speed and cost matter most. Codex is the coding agent that works across a codebase. On Google Cloud they will sit in the same interface as Gemini 4 Argon, which reached paying users earlier this month, and Anthropic's Claude. Reuters notes cloud providers usually take a cut of model sales, but no figure was given for this deal.
What it means for your business
- Don't pick a model because of your cloud; test several on your own tasks.
- Ask whether existing commitments cover new models.
- Put data terms in writing: no training use, processing region.