Gemma 3 Beginner Tutorial – Google's Most Powerful Open AI/HABR

Welcome to AI lovers! Today we are exploring Google's new Gemma 3 AI – A powerful open source model designed to work on a single GPU. If you've ever heard the hype that the Gemma 3 is “the most powerful model that can be used on a single GPU”, you're in the right place.

About Gemma 3

What is Google Gemma 3?

Think of GEMMA 3 as Google's answer to the GPT style model. open It is optimized to run anywhere, whether it's a laptop, a single server, or a high-end phone. This is the latest in the Gemma family of Google's AI models. Google has adopted the same core technology that bolsters the giant Gemini model and distilled it on the Gemma 3. Result: A set of models that you can actually download and run yourself. The Gemma 3 is all about accessibility, without sacrificing performance. In fact, the largest Gemma three models with 27 billion parameters are one of the top open AI models of quality.

Understanding the Gemma 3 Models

The Gemma 3 comes in four sizes: 1b, 4b, 12b, and 27b (a “b” with 1 billion parameters). It's bigger and generally smarter. But even a 4B or 12B can handle a lot. It is multilingual (over 140 languages ​​out of the box) and even multimodal. This means you can understand images combined with text. This is only seen on very advanced models like the GPT-4. Furthermore, Gemma 3 has expanded memory – a 128,000 tokens context window – essentially you can read and remember very long documents and conversations. For perspective, it's over 100 pages of text at a time!

Google Made Gemma 3 opens for developers. You can download the model weights for free and run them locally or call Gemma 3 via the API. If you're hosting yourself, you don't have to pay per prompt. You are licensed for commercial use with responsible AI guidelines, so you can potentially incorporate it into your product. This open model approach is similar to Meta's Llama 2, but Google is moving further with multimodal and high efficiency.

You can use your favorite tools Hugging face trances, oramas, Pitorch, Google AI Edge, Unsloth, Vllm and more. In this tutorial, we will show you how to use it by hugging your face. At the end of the tutorial, you will find a web application similar to ChatGpt running locally.

Key Features of Gemma 3

• Small but strong. Gemma 3 is powerful. Even the Big 27B model can run on one GPU. It's faster and smaller than many other AIs.

• Read long text. You can read and understand very long documents, such as books and contracts, at once.

• I understand photography and text. You can view images and ask questions. I know how to look at the photos and give you a competent answer.

• Speaks over 140 languages: Gemma works in many languages, including Spanish, French, and Japanese.

• Provides clean data. It can reply in a JSON-like format, making it easy for developers to use the answers in their apps.

• It runs on a small device. Smaller versions are available so you can run them on laptops with low memory.

• Built-in safety. Google added the ShieldGemma 2 tool to keep it safe and block harmful content.

Minimum requirements for running Gemma 3 locally

If you are running gemma 3 locally on your computer, make sure your device meets the minimum requirements.

How to access Gemma 3

There are multiple ways to work with this model. In this tutorial, we will show you three ways:

• Google AI Studio

• Hugging my face

• Google API

How to use Gemma 3 in Google AI Studio

• Open this link in your browser: https://aistudio.google.com/

• To use AI Studio, you need a Google account.

• click “Create a Prompt”.

• in Model Settings,choice Gemma 3 (You may see options such as “Gemma 27B” or “Gemma 4B”.)

• Writing, coding, questions, whatever Gemma wants to help you enter.

• Click “run” Button to get responses from Gemma 3.

How to use Gemma 3 with a hugging face trance (Python)

If you're happy with Python, this is easy. The hugging face has three Gemma models on the hub. All you do is install the Trans Library and then download the model. I don't go into every step in detail, but you can read my full tutorial on hugging your face.

Before you begin, you need to:

• Install like vscode like IDE

• Install the Python language

Next, follow these steps:

• Open https://huggingface.co and create an account there.

• Click the Models link in the top NAV, enter Gemma, and select one of Google's models from the list to filter the results.

Next, open the terminal and paste these commands one by one.

python3 -m venv venv
source venv/bin/activate
pip install transformers datasets evaluate accelerate
pip install torch

Next, open the IDE and create a new file. main.pyusing the following code:

from transformers import pipeline
import torch

pipe = pipeline(
    "image-text-to-text",
    model="google/gemma-3-4b-it",
    device="cpu",
    torch_dtype=torch.bfloat16
)

messages = [
    {
        "role": "system",
        "content": [{"type": "text", "text": "You are a helpful assistant."}]
    },
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    }
]


output = pipe(text=messages, max_new_tokens=200)
print(output[0]["generated_text"][-1]["content"])

Use this command to run your Python code.

python3 main.py

In this example, we asked Gemma to look at the image at the URL and identify the candy animal. I used the Gemma 3-4B model and device-CPU.

Complete code with other examples you can find in my git repository – gemma3

How to use Gemma 3 via Google's API

Before you begin, like in the example of a hug face, you will need to:

• Install like vscode like IDE

• Install the Python language

Next, follow these steps:

• Open https://aistudio.google.com/ and if you don't have one, create a Google account.

• next,[APIキーを取得]Click the button. This page requires that you generate an API key. Required to connect to the Google Cloud API.

Next, open the terminal and paste these commands one by one.

pip install google-generativeai
pip install google-genai

Next, open the IDE and create a new file. For example, create the following code using Google-Cloud-Api-Terminal.py:

import google.generativeai as genai

# Configure the API key
genai.configure(api_key="[REPLACE-THIS-TEXT-WITH-YOUR-API-KEY]")

# Initialize the GenerativeModel with the model name as a positional argument
model = genai.GenerativeModel("gemma-3-27b-it")

# Generate content
response = model.generate_content("What is the latest version of iphone?")

# Print the generated text
print(response.text)

In this example, we used 27 billion parameter models and a simple text prompt.

Complete code with other examples you can find in my git repository – gemma3

A simple web application similar to ChatGpt using Gemma 3

So let's create a simple web application that can be collected together and run locally on your computer. In this example, we use the Google Cloud API, but you can hug your face and download the model to your computer.

Use Chain Light Library. ChainLit is an open source Python package for building production-ready conversational AI.

Before you begin, you need to:

Next, open the terminal and paste these commands one by one.

pip install google-generativeai
pip install google-genai
pip install chainlit

Next, open the IDE and create a new file. For example, create Google-Cloud-Web-interface.py with the following code:

import chainlit as cl
import google.generativeai as genai

# Configure the Generative AI client
genai.configure(api_key="[REPLACE-THIS-TEXT-WITH-YOUR-API-KEY]")

# Initialize the GenerativeModel
model = genai.GenerativeModel("gemma-3-27b-it")

@cl.on_message
async def handle_message(message: cl.Message):
    # Generate content based on the user's input
    response = model.generate_content(contents=[message.content])
    reply = response.text

    # Send the generated response back to the user
    await cl.Message(content=reply).send()

Use this command to run your Python code.

chainlit run google-cloud-web-interface.py

If everything is fine, the web application will appear at http://localhost:8000/.

Video Tutorial

There is a video tutorial that explains everything in detail. You can find it on YouTube.

Check out YouTube: Gemma 3 Full Guide

Conclusion

In conclusion, Google Gemma 3 is a very capable AI that can run locally and customize itself. Whether you're refined it for custom apps or for enthusiasts exploring AI on PC, Gemma 3 offers immeasurable possibilities.

If you think this tutorial is helpful, Please give me a thumbs up and Subscribe More cutting-edge AI content. If you have any questions or do something cool with Gemma 3, please drop a comment. I'd like to hear it.

Thank you for reading, and until next time, Happy Coding! 👋

cheers! 😉

SRC: Gemma 3 Tutorial