Google Cloud is expanding its offering with models from OpenAI. According to The Decoder, it is thereby the first major cloud provider to offer all three leading model families on one platform: Gemini, Claude, and the GPT models. For companies building AI applications, this can considerably simplify both selection and operation.
What happened
Google Cloud has announced that OpenAI's GPT models will now also become available on the platform. Until now, the platform offered Google's in-house Gemini models and Anthropic's Claude models. With the GPT models, all three families are now under one roof.
Three models are launching initially:
- GPT-6 Astra
- GPT-6.1 Sol
- GPT-6 Luna
For now, they are accessible only as a private preview for enterprise customers. This means it is not yet a general launch for everyone. Interested companies should expect that access will initially be limited and tied to certain conditions.
OpenAI itself has also presented the partnership. It states that so-called frontier models and Codex workflows are coming to Google Cloud's Model Garden. They are supposed to be available via the Google Cloud Marketplace. The Model Garden is the catalog in which Google bundles various models, while the Marketplace is the route through which companies can obtain offerings within their existing cloud environment.
The three models at a glance
Only a few details can be derived from the available information. The following table summarizes what is known:
| Model | Status at launch | Target group |
|---|---|---|
| GPT-6 Astra | Private preview | Enterprise customers |
| GPT-6.1 Sol | Private preview | Enterprise customers |
| GPT-6 Luna | Private preview | Enterprise customers |
What each of the individual models is best suited for is not stated in the sources. Anyone who wants to make a decision should wait for technical descriptions and run their own tests.
Background
To put the news in context, it helps to look at how the market for AI models is structured. The major model developers usually make their systems available in two ways: directly through their own interfaces and products, or through the platforms of large cloud providers. For companies, the second route is often attractive because they already store data, run applications, and manage access rights there anyway.
For this reason, cloud providers have for some time been pursuing a kind of supermarket strategy: they want to offer as many models as possible, so that customers do not have to switch to another provider if they want to use a particular model. Google has its own top-tier model in Gemini, but had already opened the platform to models from other manufacturers. With the OpenAI models, a gap is now closing that many customers are likely to have felt.
Why model variety matters for companies
No model is the best at every task. Some deliver convincing results with text, others are stronger at programming, analyzing long documents, or processing images. Models also differ in speed and operating costs. Companies that run several applications therefore often use more than one model.
Until now, this often meant extra effort: separate contracts, separate access, different billing, and separate security reviews. If all three families can be reached through one platform, this effort can decrease. Whether this really works so smoothly in practice, however, remains to be seen.
The role of Codex
In its announcement, OpenAI explicitly mentions Codex workflows. At OpenAI, Codex stands for tools centered on programming with AI. "Workflow" means that not just individual questions and answers are run, but entire work processes, such as writing, checking, and revising code. The fact that such processes are to become usable in the Google Cloud environment in the future is of particular interest to developer teams.
- Claude Duet: Anthropic Launches a Proactive AI Agent
- Anthropic gives METR employee-like access
- Microsoft: AI Agents as an Insider Risk in Companies
- GPT-6 is here – and ChatGPT now answers with interface controls
- OpenAI Decisions API: you only pay for the input
What this means
The practical consequences differ depending on the audience. Since the models are only in private preview, the points below should be read as a cautious assessment, not as a reliable forecast.
For developers
Those already working on Google Cloud could in future compare different models within the same environment. This makes A/B testing easier: the same task is solved with Gemini, Claude and a GPT model, and the team sees which result fits best in terms of quality, speed and stability.
It makes sense to build applications so that the model can be swapped out. A clear layer between your own software and the model prevents a later switch from requiring a major overhaul.
For small and medium-sized businesses
Small companies rarely have their own AI team. For them, a central platform can make getting started easier, because they don't have to negotiate with several providers individually. At the same time: the private preview is aimed at enterprise customers, and it is unclear how broad access actually is at the outset.
Before making a decision, it is worth clarifying a few questions:
- Which tasks should be automated, and how important is quality in each individual case?
- Where is data processed, and does this fit your own data protection requirements?
- How are access rights and billing handled when several models are used?
- Is there a plan in case a model later becomes unavailable or changes?
For private individuals
Ordinary users will notice little directly from this announcement at first, since it is aimed at businesses. Indirectly, however, it can have consequences: if companies can switch between models more easily, this can intensify competition among providers. In the long term, this may affect the quality of apps and services that rely on such models in the background.
Open questions
Based on the information known so far, several points remain unclear:
- When the models will become generally available beyond the private preview.
- How the three models differ from one another and what they are intended for.
- What conditions apply to businesses in the preview.
Anyone following developments should therefore keep an eye on the official announcements from Google Cloud and OpenAI, and regard statements about features as reliable only after checking them for themselves.
Frequently asked questions
What did Google Cloud announce?
Google Cloud is the first major cloud provider to offer all three leading model families: Gemini, Claude and now also OpenAI's GPT models.
Which OpenAI models are included at launch?
GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are available at launch.
Who can use the OpenAI models on Google Cloud?
At first, there is a private preview for enterprise customers. General access is not mentioned in the source material.
Where will the models be available, according to OpenAI?
According to OpenAI, the models and Codex workflows are set to appear in Google Cloud's Model Garden and to be available through the Google Cloud Marketplace.
