Skip to main content

Google Vertex AI

Status: ✅ Supported

Google Cloud Vertex AI provides access to Gemini models.

Supported Models​

Model IDNameContext
deepseek-ai/deepseek-v3.1-maasDeepSeek V3.1163K
deepseek-ai/deepseek-v3.2-maasDeepSeek V3.2163K
zai-org/glm-4.7-maasGLM-4.7200K
zai-org/glm-5-maasGLM-5202K
openai/gpt-oss-120b-maasGPT OSS 120B131K
openai/gpt-oss-20b-maasGPT OSS 20B131K
gemini-2.0-flashGemini 2.0 Flash1M
gemini-2.0-flash-liteGemini 2.0 Flash Lite1M
gemini-2.5-flashGemini 2.5 Flash1M
gemini-2.5-flash-liteGemini 2.5 Flash Lite1M
gemini-2.5-flash-lite-preview-06-17Gemini 2.5 Flash Lite Preview 06-1765K
gemini-2.5-flash-lite-preview-09-2025Gemini 2.5 Flash Lite Preview 09-251M
gemini-2.5-flash-preview-04-17Gemini 2.5 Flash Preview 04-171M
gemini-2.5-flash-preview-05-20Gemini 2.5 Flash Preview 05-201M
gemini-2.5-flash-preview-09-2025Gemini 2.5 Flash Preview 09-251M
gemini-2.5-proGemini 2.5 Pro1M
gemini-2.5-pro-preview-05-06Gemini 2.5 Pro Preview 05-061M
gemini-2.5-pro-preview-06-05Gemini 2.5 Pro Preview 06-051M
gemini-3-flash-previewGemini 3 Flash Preview1M
gemini-3-pro-previewGemini 3 Pro Preview1M
gemini-3.1-pro-previewGemini 3.1 Pro Preview1M
gemini-3.1-pro-preview-customtoolsGemini 3.1 Pro Preview Custom Tools1M
gemini-embedding-001Gemini Embedding 0012K
gemini-flash-latestGemini Flash Latest1M
gemini-flash-lite-latestGemini Flash-Lite Latest1M
moonshotai/kimi-k2-thinking-maasKimi K2 Thinking262K
meta/llama-3.3-70b-instruct-maasLlama 3.3 70B Instruct128K
meta/llama-4-maverick-17b-128e-instruct-maasLlama 4 Maverick 17B 128E Instruct524K
qwen/qwen3-235b-a22b-instruct-2507-maasQwen3 235B A22B Instruct262K

Setup​

1. Google Cloud Setup​

  1. Create or select a Google Cloud project
  2. Enable the Vertex AI API: 
    gcloud services enable aiplatform.googleapis.com
  3. Ensure billing is enabled

2. Authentication​

Clawrium uses Application Default Credentials (ADC). Set up authentication:

# Install gcloud CLI if not already installed
# https://cloud.google.com/sdk/docs/install

# Authenticate
gcloud auth application-default login

Or use a service account:

# Create service account
gcloud iam service-accounts create clawrium-provider \
  --display-name="Clawrium Provider"

# Grant Vertex AI User role
gcloud projects add-iam-policy-binding PROJECT_ID \
  --member="serviceAccount:clawrium-provider@PROJECT_ID.iam.gserviceaccount.com" \
  --role="roles/aiplatform.user"

# Create and download key
gcloud iam service-accounts keys create key.json \
  --iam-account=clawrium-provider@PROJECT_ID.iam.gserviceaccount.com

# Set environment variable
export GOOGLE_APPLICATION_CREDENTIALS=/path/to/key.json

3. Add to Clawrium​

clm provider add my-vertex --type vertex

Note: Vertex AI uses Google Cloud authentication, not an API key.

4. Select Model​

Choose a default model during setup:

  • gemini-2.5-pro (best quality)
  • gemini-2.5-flash (recommended balance)

Configuration​

# View provider details
clm provider list

# Change default model
clm provider edit my-vertex --model gemini-2.5-flash

# Remove provider
clm provider remove my-vertex

Pricing​

Vertex AI uses pay-per-use pricing. Check Vertex AI pricing for current rates.

Approximate costs:

  • Gemini 2.5 Pro: ~$1.25/1M input tokens, ~$10/1M output tokens
  • Gemini 2.5 Flash: ~$0.15/1M input tokens, ~$0.60/1M output tokens

Benefits​

  • Google Cloud integration: Works with GCP services
  • Enterprise features: Fine-tuning, batch prediction
  • Global infrastructure: Low latency worldwide
  • Gemini models: Google's most capable models

Usage in Agents​

During agent onboarding:

clm agent configure my-agent
# Select my-vertex when prompted for provider

Troubleshooting​

"Permission denied"

  • Verify Vertex AI API is enabled
  • Check IAM permissions (needs aiplatform.user)
  • Ensure billing is enabled

"Authentication failed"

  • Run gcloud auth application-default login
  • Check GOOGLE_APPLICATION_CREDENTIALS is set correctly
  • Verify service account has proper roles

"Model not found"

  • Check your region supports the model
  • Verify the model name is correct
  • Some models may be in preview/limited availability

Back to Providers