OpenAI GPT-6 Models Enter Google Cloud Model Garden in Private Preview

OpenAI and Google Cloud announced on October 21, 2026 that three GPT-6 models have entered private preview inside Google Cloud Model Garden. The move gives select enterprise customers API access to GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna without a separate OpenAI contract. Models run on dedicated Google Cloud infrastructure and appear on existing GCP invoices.

Model lineup and preview access

GPT-6 Astra serves as the flagship model for complex coding, computer-use tasks, scientific research and professional workflows. GPT-6.1 Sol delivers near-Astra performance at lower cost for agentic coding and professional work. GPT-6 Luna targets high-volume, everyday tasks where speed and cost efficiency matter most. These models reached public release in September 2026 and now extend to Google Cloud customers through an interest form.

Access occurs via Model Garden APIs with rollout already underway for approved accounts. General availability is scheduled for the coming weeks. Codex CLI and Desktop workflows can already route inference requests to the new endpoints. Planned additions include deeper integration with Google Antigravity coding tools and Gemini Agent for business users.

Google Cloud now hosts the three leading closed-source families—Gemini, Anthropic Claude and OpenAI GPT-6—plus more than 200 additional models in a single catalog. Microsoft remains OpenAI’s primary cloud partner, yet the Google Cloud arrangement adds a second major distribution channel alongside existing AWS availability.

Enterprise controls and procurement

Billing flows through Google Cloud Marketplace so usage counts against existing spend commitments and Flexible Savings Plans. Prompts, context and outputs remain inside the customer data perimeter and are not used for model training. Infrastructure isolation includes VPC Service Controls, unified audit logging and authenticated identities. Zero Data Retention and Confidential Compute features are listed for future release.

The arrangement supports sovereign cloud inference requirements by keeping data within Google Cloud regions chosen by the customer. No additional OpenAI procurement process is required for eligible accounts. This single-invoice model reduces administrative overhead compared with managing separate vendor contracts and usage reports.

Enterprises evaluating multi-model strategies can now test GPT-6 performance alongside Gemini and Claude without leaving the Google Cloud console. Early users report simplified governance because security policies, logging and cost allocation remain consistent across providers. Specific pricing, supported regions and the exact general-availability date have not yet been published.

Next steps for customers and platform expansion

Broader rollout will expand model availability and add the listed integrations into Antigravity and Gemini tools. Customers interested in private-preview access should submit the Google Cloud interest form and prepare existing spend commitments for Marketplace transactions. Teams already using Codex workflows can begin routing requests to the new endpoints immediately.

The addition strengthens Google Cloud’s position as a neutral platform for frontier models while preserving OpenAI’s direct relationship with Microsoft. Organizations that require both high performance and strict data governance now have an additional verified route for GPT-6 deployment. Further updates on regional coverage and new feature enablement are expected as the preview expands.