Stop 16 · AI APIs & Cloud AI PlatformsTopics · 8/9

Google Vertex AI (Overview)

In short

The Gemini API is great for a side project. When a company on Google Cloud wants Gemini plus security, pipelines and model hosting in one place, they use Vertex AI.

Google Cloud's unified ML platform — covers training, deployment and serving of both custom models and foundation models like Gemini.

The blob holds a single spanner while standing in front of a big tidy garage full of tools and a car lift.

Think of it like this

A single spanner vs the whole garage

A spanner fixes one bolt. A full garage has lifts, tools, spare parts, a billing desk and a manager. The Gemini API is the spanner. Vertex AI is the garage: models, training, hosting, pipelines and search, all inside your Google Cloud project.

Watch it happen

What lives inside Vertex AI

Vertex AI

Model Garden

Gemini models

Open models

Training

Endpoints

Pipelines

1/5Vertex AI is Google Cloud's ML platform. Everything runs inside your GCP project with its IAM and billing.

The big ideas

What you need to know

Projects and IAM

Auth is Google Cloud IAM, not a simple API key. Locally you log in with gcloud and use your project.

Same SDK, one flag

The google-genai SDK can talk to Vertex by setting vertexai=True with a project and location.

Location is part of the call

You pick a region or the global endpoint. Model availability can differ by location.

Classic ML too

Vertex AI also hosts your own scikit-learn or PyTorch models, not just Gemini.

See it in code

Walk through the code

python

1# First, once: gcloud auth application-default login2from google import genai3 4client = genai.Client(5    vertexai=True,6    project="my-gcp-project",7    location="us-central1",8)9response = client.models.generate_content(10    model="gemini-2.5-flash",11    contents="Give 3 tips for a first ML internship.",12)13print(response.text)

1/3Log in once so the SDK can use your Google Cloud credentials.

Calling Gemini through Vertex AI. Same SDK as the Gemini API, but auth and billing go through your Google Cloud project.

Save yourself the pain

Common mistakes

Oops

Thinking Vertex AI and the Gemini API are two different model families.

Do this instead

They can serve the same Gemini models. The difference is auth, billing, governance and the extra platform features.

Oops

Experimenting on a cloud account with no budget alert.

Do this instead

Set a billing budget and alert in Google Cloud before you try anything, and clean up afterwards.

Your mission

Compare the two doors

~25 min

Quick check

Your startup has one developer and wants to test Gemini this weekend. Best starting point?

Go deeper

1Read Vertex AI's overview to understand its scope beyond just LLM API access.
2Learn how Vertex AI Model Garden compares to calling the Gemini API directly.
3Understand where Vertex AI fits for teams doing both custom ML and LLM work.