OpenAI announced on 21 October 2026 that its frontier models and Codex agentic workflows are coming to Google Cloud's Model Garden, the same catalogue that serves Google's own Gemini models S¹. Google Cloud confirmed the move in its own blog post the same day P⁴.
My read: This is the first time I've seen OpenAI voluntarily place its models on a rival cloud's first-party model catalogue rather than just offering an API endpoint. The announcement says "frontier models" without naming a single one, and gives no pricing or launch date. I'd watch whether GPT-4o or the o-series actually appear, or whether this is limited to the open-weight gpt-oss models that OpenAI already publishes on GitHub under an Apache licence P³.
The deal covers two Google Cloud surfaces. The Model Garden is Google's curated library of deployable AI models, and the Google Cloud Marketplace is where enterprise customers already buy third-party software S¹. The Vertex Model Garden SDK lets developers deploy open models and export model weights through Python P⁵. Codex workflows are part of the package: OpenAI has been developing Codex as an agentic coding system. Bringing those workflows into Google's infrastructure means a developer building on Google Cloud could eventually run OpenAI's coding agents without leaving the Google ecosystem.
Neither company names the models, sets prices, or commits to a date
The announcement does not name specific models, disclose pricing, or give a launch date S¹. No one outside the two companies has confirmed which frontier models will appear — this has not been independently verified. The word "coming" signals that nothing is live yet.
Google Cloud's blog presents the move as part of a broader effort to let customers pick among different models and infrastructure options P⁴ — that framing comes from the vendor's own post, not an independent assessment. The technical integration details remain unspecified.
For a platform team running on Google Cloud that has been calling OpenAI's API directly, this could eventually mean routing those requests through Vertex AI instead, consolidating billing and governance under one cloud provider. A fintech building compliance tooling on Google Cloud, for instance, could deploy OpenAI models alongside Gemini through a single identity and audit framework rather than managing separate vendor relationships. That convenience only materialises when pricing and availability land.
OpenAI's open-weight models, gpt-oss-120b and gpt-oss-20b, already exist on GitHub under an Apache 2.0 licence with 20,454 stars P³ — a platform figure not independently verified by this article. Whether those are the "frontier models" headed to Model Garden, or whether the deal includes proprietary models, is the question the announcement deliberately leaves open. Both companies have said only that the models are "coming," with no further date given S¹.
Sources: S1 — OpenAI models are coming to Google Cloud · P2 — OpenAI models are coming to Google Cloud · P3 — openai/gpt-oss · P4 — Google Cloud brings OpenAI models to customers · P5 — vertexai/model_garden/README.md
Written from 5 sourced items, 5 of them primary.
