OpenAI on Google Cloud

OpenAI models are available on the Gemini platform through Google Cloud, with usage billed to your Google Cloud account. The models run within Google Cloud’s security perimeter, and your organization manages access through Google Cloud IAM.

This guide covers how to make a Responses API request through Google Cloud and how feature support differs from the first-party OpenAI API.

The following models are available:

Before you begin

Before you can make requests, your Google Cloud administrator needs to:

• Enable billing for the project, arrange project approval, and coordinate acceptance of the Marketplace private offer with your billing administrator.
• Enable the Vertex AI API (aiplatform.googleapis.com) and the model in Model Garden, including acceptance of the model terms.
• If your organization’s policy restricts model access, allow the model in that policy.
• Grant the user or service account making requests permission to call the model, for example by assigning the roles/aiplatform.user role.

Once the project is ready, ask your administrator for the project ID.

Make your first API call

For local development, install the Google Cloud CLI and sign in with the Google account that has access to your project:

gcloud auth application-default login

Get an access token in the terminal where you’ll run the request:

export GOOGLE_CLOUD_ACCESS_TOKEN="$(gcloud auth application-default print-access-token)"

When the access token expires, run this command again to get a new one using your saved sign-in.

Send the request with cURL or the OpenAI SDK, replacing your-project-id with your project ID.

1
2
3
4
5
6
7
8
9
10
11PROJECT_ID="your-project-id"
PROJECT_URL="https://aiplatform.googleapis.com/v1/projects/$PROJECT_ID"

curl --fail-with-body "$PROJECT_URL/locations/global/publishers/openai/v1/responses" \
  -H "Authorization: Bearer $GOOGLE_CLOUD_ACCESS_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-6-luna",
    "input": "Write a one-sentence bedtime story about a unicorn.",
    "store": false
  }'

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16import OpenAI from "openai";

const projectId = "your-project-id";

const client = new OpenAI({
  apiKey: process.env.GOOGLE_CLOUD_ACCESS_TOKEN,
  baseURL: `https://aiplatform.googleapis.com/v1/projects/${projectId}/locations/global/publishers/openai/v1`,
});

const response = await client.responses.create({
  model: "gpt-6-luna",
  input: "Write a one-sentence bedtime story about a unicorn.",
  store: false,
});

console.log(response.output_text);

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19import os

from openai import OpenAI

project_id = "your-project-id"

client = OpenAI(
    api_key=os.environ["GOOGLE_CLOUD_ACCESS_TOKEN"],
    base_url=(
        f"https://aiplatform.googleapis.com/v1/projects/{project_id}"
        "/locations/global/publishers/openai/v1"
    ),
)
response = client.responses.create(
    model="gpt-6-luna",
    input="Write a one-sentence bedtime story about a unicorn.",
    store=False,
)
print(response.output_text)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32package main

import (
	"context"
	"fmt"
	"os"

	"github.com/openai/openai-go/v3"
	"github.com/openai/openai-go/v3/option"
	"github.com/openai/openai-go/v3/responses"
)

func main() {
	ctx := context.Background()
	projectID := "your-project-id"
	client := openai.NewClient(
		option.WithAPIKey(os.Getenv("GOOGLE_CLOUD_ACCESS_TOKEN")),
		option.WithBaseURL("https://aiplatform.googleapis.com/v1/projects/"+projectID+"/locations/global/publishers/openai/v1"),
	)

	response, err := client.Responses.New(ctx, responses.ResponseNewParams{
		Model: "gpt-6-luna",
		Input: responses.ResponseNewParamsInputUnion{
			OfString: openai.String("Write a one-sentence bedtime story about a unicorn."),
		},
		Store: openai.Bool(false),
	})
	if err != nil {
		panic(err)
	}
	fmt.Println(response.OutputText())
}

To get access tokens through Google’s authentication library instead of exporting them in your terminal, follow Google’s getting-started guide, which explains how to use Application Default Credentials (ADC) and troubleshoot API errors.

Feature availability

The table compares feature support on Google Cloud with the first-party OpenAI API.

Data retention

Your prompts and responses are not used to train OpenAI models. Default retention is up to 30 days.

If you require Zero Data Retention (ZDR), contact your OpenAI or Google Cloud representative during onboarding and wait for confirmation of support. Setting store: false on a request does not enable ZDR.