Skip to main content
Google Vertex AI provides access to Google’s Gemini models and third-party models like Claude through Google Cloud Platform. It offers enterprise-grade reliability, security, and integration with GCP services.

​
Available Models

​
Gemini 3 Series (Latest)

  • gemini-3.1-pro-preview - Most capable with advanced reasoning (1M context)
  • gemini-3-pro-preview - Advanced reasoning and thinking
  • gemini-3-flash-preview - Fast with thinking support

​
Gemini 2 Series

  • gemini-2.5-pro - Most capable Gemini 2.5 (1M context)
  • gemini-2.5-flash - Fast and efficient (1M context)
  • gemini-2.0-flash - Fast and versatile (1M context)

​
Gemini 1.5 Series

  • gemini-1.5-pro - Capable and reliable (1M context)
  • gemini-1.5-flash - Fast and efficient (1M context)
  • gemini-1.5-flash-8b - Compact and efficient (1M context)
All Gemini models support:
  • Massive 1M token context windows
  • Multimodal input (text + images)
  • Tool calling and parallel execution
  • Thinking/reasoning (Gemini 3 series)

​
Prerequisites

Before configuring Vertex AI in Forge:
  1. Google Cloud Account: Active GCP account with billing enabled
  2. GCP Project: Project with Vertex AI API enabled
  3. Authentication: Google Cloud CLI installed and configured

​
Setup Steps

1

Install Google Cloud CLI

If not already installed:macOS:
brew install google-cloud-sdk
Linux:
curl https://sdk.cloud.google.com | bash
exec -l $SHELL
Windows: Download from Google Cloud SDK
2

Authenticate and Configure GCP

Set up your GCP credentials:
# Authenticate with Google
gcloud auth login

# Set your project ID
gcloud config set project YOUR_PROJECT_ID

# Enable Vertex AI API (if not already enabled)
gcloud services enable aiplatform.googleapis.com
3

Configure Application Default Credentials

For Forge to access Vertex AI, set up ADC:
gcloud auth application-default login
This creates credentials that Forge can use automatically.
4

Configure Forge

Run the interactive login command:
forge provider login
Select Vertex AI and provide:
  • Project ID: Your GCP project ID
  • Location: GCP region (e.g., us-central1 or global)
  • Auth Method: Choose “Google ADC” (recommended)
5

Select a Model

Set your default model in forge.yaml:
model: gemini-2.5-pro
6

Verify Connection

Start Forge and test:
forge
Try a prompt:
> What are the key features of Vertex AI?

​
Configuration

​
Required Parameters

  • PROJECT_ID: Your Google Cloud project ID
  • LOCATION: GCP region (e.g., us-central1, europe-west1, or global)

​
API Endpoints

The endpoint format varies by location: Global location: 
https://aiplatform.googleapis.com/v1beta1/projects/{PROJECT_ID}/locations/global/publishers/google
Regional location: 
https://{LOCATION}-aiplatform.googleapis.com/v1beta1/projects/{PROJECT_ID}/locations/{LOCATION}/publishers/google

​
Authentication Methods

Forge automatically uses ADC when configured with “Google ADC” method:
  • Tokens are refreshed automatically
  • No manual token management needed
  • Works seamlessly with GCP services

​
Manual API Token

You can also provide a token manually: 
# Get access token
gcloud auth print-access-token

# Use with forge provider login
forge provider login
# Select Vertex AI
# Choose "API Key" method
# Paste the token
Manual tokens expire after 1 hour. Use Google ADC for automatic token refresh.

​
Model Selection

​
For Maximum Context

All Gemini models support 1M context:
  • gemini-3.1-pro-preview - Best overall
  • gemini-2.5-pro - Excellent capability
  • gemini-1.5-pro - Reliable choice

​
For Speed

  • gemini-3-flash-preview - Fast with thinking
  • gemini-2.5-flash - Fast and capable
  • gemini-1.5-flash - Quick responses
  • gemini-1.5-flash-8b - Ultra-fast

​
For Reasoning

Gemini 3 models support extended thinking:
  • gemini-3.1-pro-preview - Advanced reasoning
  • gemini-3-pro-preview - Strong reasoning
  • gemini-3-flash-preview - Fast reasoning

​
Switching Models

Change models during a session: 
/model gemini-3.1-pro-preview

​
Regions and Availability

  • us-central1 - US Central (Iowa)
  • us-east4 - US East (Virginia)
  • europe-west1 - Europe (Belgium)
  • asia-northeast1 - Asia (Tokyo)
  • global - Global endpoint (auto-routing)

​
Choosing a Region

Use global if:
  • You want automatic routing
  • Latency is not critical
  • You don’t need regional data residency
Use specific region if:
  • You need low latency
  • Compliance requires data residency
  • You’re using other regional GCP services

​
Features

​
Massive Context Windows

Gemini models support 1M tokens:
  • Process entire codebases
  • Analyze large documents
  • Long conversation history
  • Complex multi-file operations

​
Multimodal Capabilities

  • Image understanding
  • Diagram analysis
  • Screenshot interpretation
  • Combined text and visual reasoning

​
Thinking Mode

Gemini 3 models show reasoning:
  • Explicit thought process
  • Step-by-step logic
  • Problem decomposition
  • Self-verification

​
Enterprise Features

  • Audit Logging: Full request/response logging
  • VPC Service Controls: Network security
  • Customer-Managed Keys: Data encryption
  • SLA: 99.9% uptime guarantee

​
Best Practices

​
Authentication

Never commit GCP credentials to version control. Always use ADC or service accounts.
For Development:
  • Use gcloud auth application-default login
  • Let Forge automatically refresh tokens
For Production:
  • Use service accounts with minimal permissions
  • Rotate credentials regularly
  • Enable audit logging

​
Cost Management

Model Selection:
  • Use Flash models for simple tasks (lower cost)
  • Use Pro models for complex reasoning (higher cost)
  • Monitor usage in GCP Console
Token Optimization:
  • Use smaller context when possible
  • Cache common prompts
  • Batch similar requests

​
Rate Limits

Vertex AI enforces quotas:
  • Requests per minute: Varies by model and region
  • Tokens per minute: Varies by model
Check quotas in GCP Console.

​
Troubleshooting

​
Authentication Errors

If authentication fails: 
# Re-authenticate
gcloud auth application-default login

# Verify credentials
gcloud auth application-default print-access-token

# Check project
gcloud config get-value project

​
API Not Enabled

If you see “API not enabled”: 
# Enable Vertex AI API
gcloud services enable aiplatform.googleapis.com

# Verify it's enabled
gcloud services list --enabled | grep aiplatform

​
Permission Denied

If you lack permissions:
  1. Check IAM roles in GCP Console
  2. Ensure you have Vertex AI User role
  3. Contact your GCP admin for access

​
Region Not Available

If a model isn’t available in your region:
  1. Try the global location
  2. Check model availability
  3. Switch to an available region

​
Token Expiration

If using manual tokens and they expire: 
# Get new token
gcloud auth print-access-token

# Or switch to ADC
forge provider login
# Select Vertex AI
# Choose "Google ADC"

​
Deprecated: Environment Variable Setup

Using environment variables is deprecated. Please use forge provider login instead.
For backward compatibility: 
# .env
PROJECT_ID=your-project-id
LOCATION=us-central1
VERTEX_AI_AUTH_TOKEN=your-token

# forge.yaml  
model: gemini-2.5-pro

​
Next Steps

Build docs developers (and LLMs) love

Get started for freeTalk to us