# OpenAI brings GPT-6 Astra, Sol, and Luna to Google Cloud with unified billing and Codex integration

Enterprise customers can deploy OpenAI's frontier models through Google Cloud Marketplace, drawing down existing commitments on a single invoice.

- Canonical: https://reviewblip.com/news/openai-brings-gpt-6-astra-sol-and-luna-to-google-cloud-with-unified-billing-and-codex-integration/
- Published: 2026-10-21T15:04:58.671Z
- Updated: 2026-10-21T15:04:58.671Z
- Author: ReviewBlip
- Category: Cloud & Enterprise AI
- Type: News

![Modern enterprise datacenter aisle flanked by high-density server racks with illuminated optical fiber data streams connecting compute nodes](https://reviewblip.com/editorial-images/a3ee018e30f95f63a63ab50958e6112f2f03d2e8d64fc8d4f357c28a3d435b8a)

Image credit: Google Cloud / OpenAI · Editorial illustration

OpenAI and Google Cloud announced on Wednesday that OpenAI's newest frontier foundation models are entering Google Cloud's Model Garden. The deployment brings three flagship architectures—the flagship reasoning model GPT-6 Astra, the cost-efficient workhorse GPT-6.1 Sol, and the high-throughput GPT-6 Luna—directly into Google's enterprise machine learning catalogue in private preview, with general enterprise availability scheduled across coming weeks.\[1\]

For corporate IT leaders, the billing mechanics represent an equally significant shift: the models are transacted directly through Google Cloud Marketplace. Enterprise consumption draws down previously committed enterprise spend agreements and counts toward unified Flexible Savings Plans alongside first-party Gemini infrastructure. Crucially for software engineering teams, developers can now bind their local Codex CLI and Desktop workflows directly into Google Cloud endpoints for enterprise inference.\[1\]

Google Cloud emphasized that it is now the only hyperscale provider hosting all three primary closed-source frontier model families alongside its internal catalog of over 200 first-party, third-party, and open-weight models. Prompts, context caches, and outputs remain strictly quarantined within each tenant's sovereign boundary and are never ingested to train foundation models, with Zero Data Retention and Confidential Compute enclaves slated for deployment by year's end.\[1\]

Our take: The arrival of OpenAI silicon endpoints on Google Cloud signals the final collapse of single-cloud generative AI exclusivity. As enterprise buyers demand unified billing, zero-data-retention guarantees, and portable developer tooling, artificial intelligence providers can no longer afford to lock their model weights behind a single cloud wall. Integrating native Codex workflows directly into Google Cloud removes friction for enterprise developers who prefer OpenAI's agentic tooling but remain committed to Google's data infrastructure.

## Sources
- [\[1\] Google Cloud brings OpenAI models to customers — Google Cloud Blog / WebProNews](https://www.webpronews.com/openai-models-google-cloud-model-garden-gpt-6-astra/)