Google Cloud adds three OpenAI GPT-6 models in private preview, alongside Gemini and Claude

Google Cloud began offering three OpenAI models on Oct. 21, 2026, in a private preview for select enterprise customers. According to Google Cloud, as reported by The Decoder, that makes it the only major cloud provider offering Gemini, Claude and OpenAI models on one platform.

What is available

Three OpenAI models are included at launch:

• GPT-6 Astra: positioned for demanding tasks such as coding and research.
• GPT-6.1 Sol: described as a cheaper alternative with similar performance to Astra. The performance claim comes from the announcement and has not been independently verified.
• GPT-6 Luna: a fast, cost-efficient option for everyday use.

The models will also appear in Gemini for Business and Google Antigravity.

Access and billing

Access is limited to select enterprise customers, though sign-ups are open now. All usage is billed through a single Google Cloud invoice and counts toward customers' existing Google Cloud contracts. Google says customer data will not be used for training.

What has not been disclosed

The announcement as reported does not include:

• Per-token pricing for any of the three models (pricing not yet disclosed)
• Context window sizes
• Benchmark scores
• Regional availability
• A timeline for general availability

All three models are OpenAI releases offered through Google's infrastructure. This is a distribution change, not a new model launch by Google.

A capacity tension inside Google

The Decoder notes a recurring conflict at Google: compute capacity used to serve cloud customers can strengthen the main rivals of Google's own Gemini team. The outlet says this resource conflict reportedly contributed to researcher Noam Shazeer returning to OpenAI. That account is a report, not a confirmed statement from either company.

What this means

For enterprise buyers, the practical change is procurement. Teams already committed to Google Cloud can test OpenAI, Anthropic and Google models under one contract, one bill and one data-handling policy. That lowers the cost of running side-by-side evaluations and of switching between vendors.

For Google, the strategy is to compete as the neutral platform rather than only on Gemini quality. That brings revenue from rivals' workloads, but it also directs scarce compute toward models that compete with Gemini. Whether Google can keep both businesses supplied with capacity is the open question.

The preview status limits what can be concluded yet. Without published pricing, context limits or regional availability, it is not possible to compare these offerings with OpenAI's own API or with other clouds. Those details will determine whether this is a real alternative route to OpenAI models or mainly a bundling convenience for existing Google Cloud customers.