Gemma

Gemma is Google's family of open-weight language models, derived from Gemini research but sized for local deployment and fine-tuning. Gemma 3 models range from 1B to 27B parameters and support text, image, and video input.

• Local deployment — run on consumer hardware, edge devices, and mobile
• Fine-tuning — customize for domain-specific tasks with full weight access
• Multimodal — Gemma 3 supports text, image, and video inputs
• Research — transparent architecture for academic and safety research

Developers and researchers use Gemma when they need a capable open model they can run locally, fine-tune freely, and deploy without API dependencies.

1. Download Gemma models from Hugging Face (accept the license agreement) or through Google AI.
2. Run locally with Ollama:

   ollama run gemma3:27b

3. Or use the Hugging Face Transformers library:

   pip install transformers torch

Gemma models are free and open-weight under Google's Gemma license, which permits commercial use.