Vertex AI

Platform Overview

Gemini Models

1M+ token context, multimodal (text, image, video, audio), function calling.

Model Garden

150+ models: Gemini, Claude, Llama, Mistral, and open-source models.

Grounding & RAG

Built-in Google Search grounding, Vertex AI Search for enterprise RAG.

Enterprise Security

VPC-SC, CMEK, data residency, audit logging, responsible AI controls.

Available Models

Model	Context	Best For
gemini-2.0-flash	1M tokens	Fast, multimodal, agents
gemini-2.0-pro	2M tokens	Complex reasoning, coding
claude-3-5-sonnet	200K tokens	Coding, analysis
llama-3.1-405b	128K tokens	Open model, large scale
text-embedding-005	2048 tokens	Embeddings, RAG
imagen-3	-	Image generation

Quickstart (Python)

Installation

pip install google-cloud-aiplatform

Basic Chat (Gemini)

import vertexai
from vertexai.generative_models import GenerativeModel

# Initialize
vertexai.init(project="YOUR_PROJECT_ID", location="us-central1")

# Load model
model = GenerativeModel("gemini-2.0-flash")

# Generate
response = model.generate_content("Explain quantum computing simply.")
print(response.text)

Chat with History

chat = model.start_chat()

response1 = chat.send_message("What is Python?")
print(response1.text)

response2 = chat.send_message("How does it compare to JavaScript?")
print(response2.text)

Multimodal (Vision)

from vertexai.generative_models import GenerativeModel, Part, Image

model = GenerativeModel("gemini-2.0-flash")

# From local file
image = Image.load_from_file("image.jpg")

# Or from GCS
# image = Part.from_uri("gs://bucket/image.jpg", mime_type="image/jpeg")

response = model.generate_content([
    "What's in this image?",
    image
])
print(response.text)

Function Calling (Tools)

from vertexai.generative_models import GenerativeModel, Tool, FunctionDeclaration

# Define function
get_weather = FunctionDeclaration(
    name="get_weather",
    description="Get weather for a location",
    parameters={
        "type": "object",
        "properties": {
            "location": {"type": "string", "description": "City name"}
        },
        "required": ["location"]
    }
)

# Create tool
weather_tool = Tool(function_declarations=[get_weather])

# Use with model
model = GenerativeModel("gemini-2.0-flash", tools=[weather_tool])
response = model.generate_content("What's the weather in Tokyo?")

# Check for function calls
for candidate in response.candidates:
    for part in candidate.content.parts:
        if part.function_call:
            print(f"Call: {part.function_call.name}")
            print(f"Args: {part.function_call.args}")

Embeddings

from vertexai.language_models import TextEmbeddingModel

model = TextEmbeddingModel.from_pretrained("text-embedding-005")

texts = ["Hello world", "Machine learning is powerful"]
embeddings = model.get_embeddings(texts)

for i, embedding in enumerate(embeddings):
    print(f"Text {i}: {len(embedding.values)} dimensions")

Google Search Grounding

Ground responses in real-time Google Search results to reduce hallucinations.

from vertexai.generative_models import GenerativeModel, Tool, grounding

# Enable Google Search grounding
search_tool = Tool.from_google_search_retrieval(
    grounding.GoogleSearchRetrieval()
)

model = GenerativeModel("gemini-2.0-flash", tools=[search_tool])

response = model.generate_content(
    "What are the latest AI news from this week?"
)
print(response.text)

# Access grounding metadata
for candidate in response.candidates:
    if candidate.grounding_metadata:
        for chunk in candidate.grounding_metadata.grounding_chunks:
            print(f"Source: {chunk.web.uri}")

Pricing (per 1M tokens)

Model	Input	Output
Gemini 2.0 Flash	$0.10	$0.40
Gemini 2.0 Pro	$1.25	$5.00
Claude 3.5 Sonnet	$3.00	$15.00

💡 Tip: Use gemini-2.0-flash for most workloads. It's fast, cheap, and supports 1M tokens.