Stack Explorer

Google Gemma

llm-model model

Small and open models from Google DeepMind

Official site

Supported languages

Gemma is Google's family of open language models, built with the same technology as Gemini. Available in 2B and 7B sizes, they're designed to be safe, efficient, and easy to deploy on limited hardware.

Concepts

slmdecoder-onlyinstruction-tuningsafety-tuningefficient-inference

Pros and Cons

Ventajas

  • + Backed by Google DeepMind
  • + Excellent performance for its size
  • + Permissive license for commercial use
  • + Multiple variants (base, instruct, code)
  • + Well documented and supported
  • + Optimized for efficiency

Desventajas

  • - Less capacity than larger models
  • - Less general knowledge than large LLMs
  • - Requires fine-tuning for specific tasks
  • - Limited context compared to GPT-4

Casos de Uso

  • Local device assistants
  • Mobile applications with AI
  • Efficient text processing
  • Low-cost chatbots
  • Simple code generation