Vertex AI

What is Vertex AI?

Vertex AI, marketed as the Gemini platform, is a unified cloud service for building, training, tuning, and deploying machine learning models and generative AI agents. It provides access to over 200 Google and third-party models through Agent Studio, Model Garden, and programmatic APIs. The service requires a Google Cloud account and bills based on compute, storage, and model usage.

What are the features of Vertex AI?

• Model catalog (Model Garden)

  Discover, test, customize, and deploy over 200 models, including Gemini 4 Argon, Anthropic's Claude family, and open models like Gemma.
• Agent development interface (Agent Studio)

  Design, test, and manage prompts for large generative AI models using natural language, code, images, or video inputs.
• Custom model training

  Train models with preferred machine learning frameworks, custom code, and selected hyperparameter tuning options.
• MLOps tooling

  Orchestrate workflows with Pipelines, manage versions in Model Registry, share features via Feature Store, and monitor for input skew and drift.
• Integrated notebooks

  Use Colab Enterprise or Workbench environments natively integrated with BigQuery for data and AI workloads.
• Prediction service

  Register models to Model Registry and serve batch or online predictions using prebuilt containers or custom prediction routines.

What are the use cases of Vertex AI?

• Multimodal prompt testing

  Upload images or videos to Agent Studio to generate answers, extract text from images, or convert image mockups to HTML.
• Enterprise document retrieval

  Use Agent Search to run semantic or keyword-based queries against indexed enterprise data, billed per 1,000 successful search queries.
• Production model deployment

  Move trained models from notebook code to serving infrastructure for real-world batch or online predictions.

How to use Vertex AI?

1. Initialize client

   Import the genai library and create a client instance using from google import genai and client = genai.Client().
2. Generate content

   Call client.models.generate_content() specifying a model identifier such as "gemini-2.5-flash" and passing the prompt string in the contents parameter.
3. Output response

   Print the generated text by accessing response.text from the returned object.

Vertex AI Pricing Information:

• New customer trial

  $300 in free credits valid for 90 days, applicable toward agent runtime compute, memory, and generative model queries.
• Free Tier

  Per billing account per month, includes the first 50 hours (180,000 vCPU-seconds) of compute, 100 hours (360,000 GiB-seconds) of RAM, and 1 gibibyte of storage for agent runtimes.
• Text, chat, and code generation

  Starting at $0.0001 per 1,000 characters of input and output.
• Imagen model for image generation

  Starting at $0.0001, based on image input, character input, or custom training pricing.
• Custom model training

  Priced by machine type per hour, region, and accelerators used; contact sales for estimates.
• Notebooks

  Compute and storage billed at Compute Engine and Cloud Storage rates, plus management fees based on region and instances.
• Pipelines

  Starting at $0.03 per pipeline run, plus resource and service fees.
• Vector Search

  Costs based on data size, queries per second, and number of nodes.
• Document retrieval (Agent Search)

  Approximately $4.00 per 1,000 queries for standard semantic or keyword-based retrieval, plus storage charges per GiB/month.
• Payment tiers

  Standard Pay-as-you-go (PayGo), discounted Flex PayGo for latency-tolerant jobs, premium Priority PayGo, or Provisioned Throughput subscriptions.