OpenAI's frontier models are heading to Google Cloud, a deal that puts the company's latest GPT models alongside Gemini and Claude inside Google's enterprise AI stack—and, notably, inside Antigravity, Google's agent-powered coding tool. The arrangement marks a notable expansion of OpenAI's distribution beyond its primary cloud partner, Microsoft, and gives Google Cloud customers a direct path to running GPT workloads on Google's infrastructure.

What's Actually Coming

The full suite of current GPT 6 and 6.1 models—Astra, Sol and Luna—will launch on Google Cloud, with Astra positioned as the most expensive to run of the three. The models will appear in Google's Model Garden, the catalog that already hosts more than 200 options, including Google's own Gemini family, open models such as Mistral, Llama and Gemma, and Anthropic's Claude. Google says the addition makes it the only large cloud provider offering access to all three leading closed LLMs.

OpenAI models will run entirely on Google's infrastructure, with transactions handled through the Google Cloud Marketplace. That detail matters for enterprise procurement: businesses with existing Google Cloud spending commitments can apply those commitments to GPT usage, lowering the friction of switching or adding a model provider. Google's system also supports dynamically moving workloads between LLMs based on performance, latency and efficiency, which lets customers route requests to whichever model best fits a given task.

Private Preview and Developer Access

All three OpenAI models are rolling out now in private preview for select enterprise customers. Organizations interested in early access can submit a request, which enables access through the API even for teams using Codex CLI or Codex Desktop—the inference simply runs on Google's systems rather than OpenAI's or Microsoft's. The private-preview model is a familiar pattern for Google Cloud, which has used staged rollouts to manage capacity and gather feedback before general availability.

The Antigravity piece is arguably the more interesting one for individual developers. Until now, the agent-powered coding client has supported Google's Gemini and Anthropic's Claude. With the deal, GPT models will appear in Antigravity as well. Google has not specified whether all three OpenAI models will be available there, but the same trio is the natural expectation given the Cloud deployment.

Google's Model-Agnostic Bet

Google would obviously prefer that everyone standardize on Gemini, and it recently rolled out Gemini 4 Argon across its AI footprint, positioning the flagship as back at parity with the competition. Argon does appear to be an improvement over Gemini 3.1 Pro, but it consumes tokens at an unusually high rate—a cost characteristic that gives customers a reason to look elsewhere for certain workloads. Supporting Claude and now OpenAI's models turns Google Cloud into a neutral venue for frontier AI rather than a single-vendor shop, which could help it win enterprise accounts that want optionality.

For OpenAI, the deal broadens reach without displacing Microsoft, which the company says remains its primary cloud partner. Running GPT models on Google's infrastructure is a distribution play: it meets enterprise customers where their spending commitments and data already live, and it puts OpenAI's models in front of developers inside a Google coding tool they may already be using.

What to Watch

The open questions are pricing and timing. Google has not published per-token rates for Astra, Sol or Luna on its platform, and general availability beyond the private preview has not been dated. It is also unclear whether Antigravity access will ship alongside the Cloud rollout or later. For enterprise architects, the practical test will be whether routing between Gemini, Claude and GPT models through a single cloud marketplace delivers the cost and latency benefits Google is promising—and whether the dynamic workload-routing feature works as advertised across vendors. For now, the deal signals that the frontier-model market is consolidating around multi-model cloud platforms, and that even fierce competitors see distribution as worth more than exclusivity.