OpenAI's GPT 6.1 models are coming to Google Cloud and Antigravity

OpenAI and Google have reached an arrangement that brings OpenAI's frontier models into Google Cloud services. The full GPT 6.1 lineup, Astra, Sol and Luna, will launch there, and all three are rolling out now in private preview for selected enterprise customers. Anyone wanting early access can submit a request, and access runs through the API even for users of Codex CLI or Codex Desktop.

The models will also appear in Google's Model Garden in the coming weeks, which already offers more than 200 options including Gemini, Mistral, Llama, Gemma and Claude. Inference runs on Google's infrastructure, and transactions are handled through the Google Cloud Marketplace. Google's system can move workloads between models based on performance, latency and efficiency, and Astra will be the most expensive of the three to run.

GPT models will also come to Antigravity, Google's agent-powered coding tool, which currently supports Gemini and Claude. Google has not said whether the same three models will be offered there. OpenAI says Microsoft remains its primary cloud partner.

Developers on Google Cloud who already have spending commitments can start calling GPT 6.1 models without changing their billing setup, since purchases go through the Google Cloud Marketplace.

Teams planning to use Antigravity should not assume the same GPT models will be there until Google confirms the lineup.

Because Google's system can shift workloads between models, agent builders may need to check how routing choices affect cost and latency.