Skip to content

Gemma Documentation

Google Gemma 4 Open Models

Home
Get started
Models
Run
Fine-tune
Learn
Resources

Gemma Documentation

Home
Get started
Get started
- Quickstart
- Best practices
- Text
  Text
- Multimodal
  Multimodal
- Advanced
  Advanced
Models
Models
- Overview
- Gemma 4
  Gemma 4
  - Overview
  - Model card
- Previous generations
  Previous generations
  - Gemma 3
  - Gemma 2
  - Gemma
- Variants
  Variants
  - Gemma 3n
    Gemma 3n
    
    Overview
    
    Model card
  - DiffusionGemma
    DiffusionGemma
    
    Overview
    
    Model card
    
    Diffusion explained
    
    Generate output
  - FunctionGemma
    FunctionGemma
    
    Overview
    
    Model card
    
    Formatting and best practices
    
    Function calling with HF
    
    Full sequence
    
    Fine-tuning
  - EmbeddingGemma
    EmbeddingGemma
    
    Overview
    
    EmbeddingGemma model card
    
    EmbeddingGemma 2 model card
    
    Generate embeddings
    
    Multimodal embeddings
    
    Fine-tuning
    
    Multimodal fine-tuning
  - PaliGemma
    PaliGemma
    
    Overview
    
    PaliGemma model card
    
    PaliGemma 2 model card
    
    Prompting & system instructions
    
    Generate output with Keras
    
    Fine-tuning with JAX
  - ShieldGemma
    ShieldGemma
    
    Overview
    
    ShieldGemma model card
    
    ShieldGemma 2 model card
- Research
  Research
  - RecurrentGemma
    RecurrentGemma
    
    Overview
    
    Model card
    
    Generate output with JAX
    
    Fine-tuning with JAX
  - DataGemma
  - Gemma Scope
  - Gemma APS
- Releases
Run
Run
- Overview
- Local
  Local
- Serve
  Serve
  - vLLM
  - SGLang
  - Gemini API
  - Google Cloud
- Frameworks
  Frameworks
  - LangChain
  - Unsloth
Fine-tune
Fine-tune
- Overview
- Guides
  Guides
Learn
Learn
- Overview
- Visual guides
  Visual guides
- Gemma explained
  Gemma explained
Resources
Resources
- Overview
- Demos & tools
  Demos & tools
- Legal
  Legal
  - Gemma 4 license
  - Terms of use
  - Terms archive
    Terms archive
    
    Overview
    
    April 1, 2024
    
    February 21, 2024
  - Prohibited use
  - Intended use statement

Gemma

Our most capable open models

Run Gemma 4 locally with Ollama…

Get started Try in Google AI Studio Explore models

Copyright © 2026 Google DeepMind

Change cookie settings · Made with Zensical

Cookie consent

We use cookies to recognize your repeated visits and preferences, as well as to measure the effectiveness of our documentation and whether users find what they're searching for. With your consent, you're helping us to make our documentation better.

Google Analytics

Manage settings