# Gemma on Google Cloud Google Cloud offers several routes for deploying, serving and fine-tuning Gemma 4. Pick by what you are doing, not by product name — each card jumps to the detail below. ### Serve it

Fully managed, scale-to-zero, pay-per-use GPU inference.

Your own serving cluster on TPUs or GPUs, for high throughput and low latency.

vLLM on TPUs

State-of-the-art serving performance on Cloud TPUs.

Sovereign Cloud

Extra control and compliance for sensitive workloads.

### Build with it

Agent Platform

Test and deploy straight from Model Garden in the console.

Agent Development Kit

Orchestrate agents around Gemma's reasoning and function calling.

### Train it

Training Clusters

Managed infrastructure for large-scale fine-tuning of open models.

MaxText

High-performance JAX implementation for Cloud TPUs.

## Gemini Enterprise Agent Platform [Gemini Enterprise Agent Platform](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/open-models/use-gemma) is a Google Cloud platform for rapidly building and scaling machine learning projects. Gemma 4 is available in [Model Garden](https://console.cloud.google.com/agent-platform/publishers/google/model-garden/gemma4), a curated collection of models on Gemini Enterprise Agent Platform. You can test and deploy models directly from the console. To learn more, refer to the following pages: - [Agent Platform overview](https://docs.cloud.google.com/gemini-enterprise-agent-platform/overview): Get started with Gemini Enterprise Agent Platform. - [Gemma with Gemini Enterprise Agent Platform](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/open-models/use-gemma): Use Gemma open models with Gemini Enterprise Agent Platform. ## Cloud Run [Cloud Run](https://cloud.google.com/run) is a fully managed platform to run your code or containers on top of Google's highly scalable infrastructure. [Deploy Gemma 4 on Cloud Run](https://docs.cloud.google.com/run/docs/run-gemma-on-cloud-run) using GPUs for scale-to-zero, pay-per-use inference. For larger mode sizes, [leverage advanced configurations](https://codelabs.developers.google.com/codelabs/cloud-run/cloud-run-gpu-rtx-pro-6000-gemma4-vllm){ data-model="gemma-4" } with RTX 6000 Pro GPUs and Model Streaming. ## Google Kubernetes Engine (GKE) [Google Kubernetes Engine](https://cloud.google.com/kubernetes-engine) (GKE) is a managed Kubernetes service from Google Cloud. [Run Gemma 4 on GKE](https://cloud.google.com/kubernetes-engine/docs/tutorials/serve-gemma-gpu-vllm) for enterprise-grade container orchestration. Use TPUs and GPUs to serve models with high throughput and low latency. ## Agent Development Kit (ADK) Build and orchestrate AI agents with Gemma 4 and the [Agent Development Kit (ADK)](https://adk.dev/agents/models/google-gemma/). Gemma 4's strong reasoning and function-calling capabilities make it ideal for agentic workflows. ## Gemini Enterprise Agent Platform Training Clusters [Fine-tune Gemma 4 using Gemini Enterprise Agent Platform Training Clusters](https://discuss.google.dev/t/end-to-end-guide-fine-tuning-and-serving-gemma-4-on-vertex-ai/345865). Training Clusters provides optimized infrastructure for large-scale training and fine-tuning of open models. ## vLLM with TPUs [Serve Gemma 4 on Google Cloud TPUs](https://docs.vllm.ai/projects/recipes/en/latest/Google/Gemma4.html#cloud-tpu-deployment-via-docker){ data-model="gemma-4" } for state-of-the-art serving performance. ## MaxText Gemma 4 is supported in [MaxText](https://github.com/google/maxtext), a high-performance, arbitrary-sized JAX LLM implementation for Google Cloud TPUs. ## Sovereign Cloud Gemma 4 is available on [Sovereign Cloud](https://cloud.google.com/sovereign-cloud) solutions, providing enhanced control and compliance for sensitive workloads.