Gemma
Google's lightweight open models - Gemma 3 leads open-source benchmarks
Google DeepMind · Language Models
Quick Answer
Gemma is google's lightweight open models - gemma 3 leads open-source benchmarks, made by Google DeepMind. It is free and open source. Key uses: Local AI deployment, Fine-tuning, Research.
Last updated:
What is Gemma?
Gemma 3 (released Mar 2025) is Google DeepMind's family of open-source models available in 1B, 4B, 12B, and 27B parameter sizes. The 27B model leads open-source benchmarks outperforming Llama 3.3 70B at half the size. Features 128K context and multimodal support.
What's new in Gemma in 2026?
Gemma 3 (Mar 2025) - 128K context, multimodal image input, #1 single-GPU open model; ShieldGemma 2 safety classifier; CodeGemma for coding tasks
What can you do with Gemma?
- Local AI deployment
- Fine-tuning
- Research
- Edge devices
- Multimodal apps
Pros
- Runs on consumer hardware
- Open weights
- Gemma 3 27B beats 70B models
- 128K context
- Multimodal (image input)
Cons
- Smaller capability ceiling vs frontier models
- No official fine-tuning SLAs
What are the best alternatives to Gemma?
Compare Gemma Head-to-Head
Frequently Asked Questions about Gemma
Is Gemma free?
Yes, Gemma is fully free and open source.
What is Gemma used for?
Gemma is used for: Local AI deployment, Fine-tuning, Research, Edge devices, Multimodal apps.
What are the best alternatives to Gemma?
Top alternatives to Gemma include Mistral AI, Llama, Phi-4.