Google's New Open LLM: Gemma 4
Unveiled at Google I/O 2025, Gemma 4 is the most powerful open large language model Google has shipped. You can reach it on Hugging Face, Google AI Studio and Vertex AI in 2B, 9B and 27B parameter variants.
Gemma 4 Capabilities at a Glance
- 140+ language support - including Kazakh
- 128K token window - read long documents in a single pass
- Multimodal capabilities - process both text and images
- Instruction tuned release - suited to assistant workloads
- Open license with no strings - commercial use included
Gemma 4 vs Gemini
| Feature | Gemini | Gemma 4 |
|---|---|---|
| Availability | Closed-source API | Open source |
| Deployment | Google's servers | Your own server |
| Cost | Per API call | Free (self-hosted) |
| Control | Limited | Full |
Gemma 4 exists for teams that refuse to hand over control of the model.
What Firms Build With It
Firms put Gemma 4 to work on customer-support chatbots, document summarization, RAG systems, code generation and sentiment analysis.
Gemma 4 on Google Cloud Vertex AI
Vertex AI gives you full Gemma 4 management: fine-tuning, deployment, monitoring and API access. Geosoft Cloud helps with Gemma 4 integration on Vertex AI.