OpenAI's GPT-6 models are coming to Google Cloud and Antigravity

What's happening

OpenAI and Google have reached an agreement that will make OpenAI's latest flagship models available in Google's cloud. The deal covers Model Garden, Google's catalog of more than 200 models, and Antigravity, an agent-based coding tool. According to CNBC, the launch is set for "the coming weeks." Google's own blog says a private preview for selected enterprise customers is already running.

Which models are coming

The current GPT-6 family is on offer. The descriptions come from Google:

According to Ars Technica, Astra is likely to be the most expensive to run. Google's platform can shift workloads dynamically between models based on performance, latency and cost. The idea is that customers can pick the right model for each task, with no architecture lock-in and no separate contracts.

Billing, security and availability

The OpenAI models run entirely on Google's infrastructure and are billed through the Google Cloud Marketplace. Usage appears on a single Google invoice, counts toward existing cloud commitments and qualifies for Flexible Savings Plans. For companies with a Google budget, that makes switching easy.

Google promises that prompts, context and outputs stay within the designated data region and aren't used to train foundation models. Zero data retention (no storage of data) and confidential compute (especially isolated processing) are only due "soon." Requests go through hardened, isolated gateways, every interaction runs under an authenticated identity, and VPC Service Controls and unified audit logging are included. Box named itself as a future user.

According to Google, developers can start working with the models via API right away and connect their local Codex CLI and desktop workflows to Google's Gemini platform for inference. Interested customers have to fill out a form. The models are later expected to arrive in "Gemini for business" and Antigravity as well.

Background

CNBC sees the move as part of OpenAI's strategy to reach more enterprise developers and move away from the consumer business (ChatGPT). Less than six months ago, OpenAI said Microsoft no longer held an exclusive license to its intellectual property; one day after the exclusivity ended, Amazon announced it would host OpenAI models. Now OpenAI models are available at the three largest cloud providers. Last year, OpenAI already used Google's cloud to run ChatGPT and its API.

According to CNBC, Google's cloud is growing fast: annualized revenue growth rose to around 82% in the second quarter, up from 63% in the first, compared with 37% and 42% at Amazon and Microsoft. Ramp data shows most enterprise AI spending goes to OpenAI and Anthropic. Google unveiled its Gemini 4 Argon model in September. OpenAI is reportedly approaching investors at a valuation of $1.4 trillion and aiming for an IPO.

Our take

For enterprise customers, this means more choice and less administrative overhead: if you've already committed budget to Google Cloud, you can use GPT models without adding a new vendor. For Google, it's a platform selling point, even though its own Gemini models now face competition in-house. For OpenAI, reach grows beyond Microsoft. Still unclear: when the models become generally available, what pricing looks like, and when security features such as zero data retention will actually ship. Claims like "the only hyperscaler with all three leading closed models" come from Google itself.