OpenAI Models Arrive on Google Cloud Model Garden
OpenAI and Google Cloud partner to offer frontier models and Codex workflows in Model Garden, allowing enterprise spend commitment drawdowns.
OpenAI and Google Cloud announced a commercial partnership on October 21, 2026, bringing OpenAI frontier models and Codex agentic workflows directly into Google Cloud's Model Garden with general availability planned in the coming weeks. The agreement marks a structural change in how enterprise developers procure OpenAI tooling, enabling organizations to purchase frontier systems through the Google Cloud Marketplace while drawing down existing Google Cloud Platform spend commitments.1
The transaction mechanism alters enterprise procurement math by letting corporate accounts apply contracted cloud budgets directly to OpenAI usage alongside first-party Google models. Microsoft remains OpenAI's primary cloud partner, but the deal provides Google Cloud customers direct access without establishing separate billing infrastructure.15
Integration with Google Antigravity and the Gemini Agent
In addition to Model Garden catalog availability, OpenAI models will soon integrate into Google Antigravity and the Gemini Agent. Google Cloud described the move as an effort to expand model choice and task flexibility for enterprise users.12

This expands an increasingly heterogeneous runtime inside Google Antigravity, which already hosts Claude Opus 5.5 and Claude Sonnet 5.5. For engineering teams evaluating Claude Opus 5.5 vs Sonnet 5.5 alongside Google native tooling, the addition of OpenAI endpoints provides unified API access in a single execution console.2
OpenAI shifts from backend tenant to Marketplace vendor
The commercial distribution agreement builds on an earlier infrastructure relationship. OpenAI previously finalized an agreement in May to lease Google Cloud Platform compute across the United States, Japan, the Netherlands, Norway, and the United Kingdom after prolonged negotiations.34
Until January, Microsoft Azure served as OpenAI's exclusive data center provider, before Microsoft altered the contract to maintain a right of first refusal for extra cloud workloads rather than exclusive status. To satisfy scaling needs, OpenAI expanded compute suppliers aggressively: it signed an $11.9 billion contract with CoreWeave in March for over 250,000 Nvidia GPUs across 32 data centers, and partnered with SoftBank and Oracle on the $500 billion Stargate infrastructure program.34
Standard token rates across OpenAI frontier models
Enterprise procurement planning requires mapping application workloads against standard frontier pricing. As tracked on October 13, 2026, OpenAI model rates span a wide pricing gradient across latency tiers and reasoning profiles.7
| Model | Input Per 1M Tokens | Output Per 1M Tokens |
|---|---|---|
| gpt-6-luna | $0.10 | $0.50 |
| gpt-6.1-sol | $2 | $10 |
| gpt-5.6-sol | $4 | $20 |
| chat-latest | $5 | $30 |
| gpt-6-astra | $10 | $50 |
| gpt-5.6-cyber | $12.50 | $75 |
According to pricing tracking data, for a balanced workload of 1 million input and 1 million output tokens, gpt-6-luna costs $0.60 while gpt-5.6-cyber costs $87.50, an increase of 145.8 times. Teams deploying conversational audio workflows can also balance gpt-realtime-2.1 ($4 input and $24 output per million) against gpt-realtime-2.1-mini ($0.60 input and $2.40 output per million). When designing architectures with open source AI coding agents, selecting high-volume lightweight models keeps automated token consumption manageable.7