Download Gemma 4

Access Gemma 4 models on your preferred platform. All models are released under the Apache 2.0 license and available for both research and commercial use.

HuggingFace

HuggingFace is the primary distribution platform for Gemma 4 models. Download the full-precision weights, quantized variants, and access model cards with detailed documentation.

• Gemma 4 31B Dense
• Gemma 4 26B A4B (MoE)
• Gemma 4 E4B
• Gemma 4 E2B

Ollama

The fastest way to run Gemma 4 locally. Install Ollama and pull the model with a single command:

ollama pull gemma4:31b
ollama pull gemma4:26b
ollama pull gemma4:e4b
ollama pull gemma4:e2b

Run with: ollama run gemma4:31b

Kaggle

Gemma 4 models are available on Kaggle Models with integrated notebook support for experimentation.

• Gemma 4 on Kaggle

Google AI Studio

Try Gemma 4 directly in your browser with no setup required. Get an API key for programmatic access.

• Google AI Studio

LM Studio

A desktop application for running LLMs locally with a chat interface. Download LM Studio, search for "gemma-4", and start chatting.

ModelScope (China)

For users in China, Gemma 4 models are available on ModelScope (魔搭社区) with optimized download speeds.

• Gemma 4 31B on ModelScope

GGUF Quantized Versions

Pre-quantized GGUF files are available on HuggingFace for use with llama.cpp, Ollama, LM Studio, and other compatible runtimes. Quantized versions significantly reduce memory requirements while maintaining strong performance.

Look for community-uploaded GGUF variants on HuggingFace, typically available in Q4_K_M, Q5_K_M, Q6_K, and Q8_0 quantization levels.