Gemma 3
Open SourceGoogle DeepMind · 2025
Google's lightweight open-source model family for efficient on-device and edge deployment.
Quick Facts
Parameters
2B / 7B
Context Window
8K tokens
Modalities
text, code
Open Source
Yes
License
Gemma Commercial License
Pricing
Free (open-source)
Released
2025
Developer
Google DeepMind
About
Gemma 3 is Google's lightweight open-source model family, designed for efficient deployment on consumer hardware, mobile devices, and edge environments. Available in 2B and 7B parameter configurations, Gemma 3 delivers strong performance for its size, leveraging Google DeepMind's advanced research in model compression and efficient architecture. It supports text and code generation with an 8K token context window. Gemma 3 is ideal for developers building lightweight AI applications, running inference on-device, or deploying cost-efficient AI at scale. Released under a permissive commercial license, it enables widespread adoption across research and production.
Strengths
- +Designed for efficient on-device deployment
- +Permissive commercial license
- +Strong performance for its compact size
- +Multilingual support across 100+ languages
Weaknesses
- −Limited 8K context window
- −Not suitable for complex reasoning tasks
- −Significantly less capable than larger models
Best For
On-device mobile AI applications
Edge computing and IoT deployments
Lightweight text classification and generation
Applications with strict latency and cost requirements
Pricing
Open Source
$0
- 2B and 7B model sizes
- Commercial use
- On-device deployment
Google Cloud
Pay-per-use
- Hosted API
- Vertex AI integration
- Enterprise support
Technical Specs
Parameters
2B / 7B
Context Window
8K tokens
Modalities
text, code
Languages
Open Source
Yes
License
Gemma Commercial License
Developer
Google DeepMind
Released: 2025
Related Models
Llama 4
Meta AI
Meta's next-generation open-weight LLM with MoE architecture and strong multilingual support.
Stable Diffusion 3.5
Stability AI
Open-source image generation model with local execution, full privacy, and community ecosystem.
CodeLlama 70B
Meta AI
Open-source code generation model specialized for programming tasks across multiple languages.
Mistral Large 2
Mistral AI
Mistral AI's flagship open-weight model with 123B parameters and multilingual excellence.