# Gemma 4

Google DeepMind's open model family for local, multimodal, and agentic AI workflows.

## Summary
Gemma 4 is a family of Apache 2.0 open models from Google DeepMind, designed for reasoning, multimodal inputs, edge deployments, and developer workflows that need more control than hosted-only APIs.


## Guide
Gemma 4 is Google DeepMind's latest open model family for developers who want more deployment control than a purely hosted model API can provide. It sits in the growing middle ground between frontier proprietary models and smaller local models: capable enough to test real product workflows, but available in forms that can be evaluated outside a closed chat product.

### What it is
Gemma 4 is an open model resource to evaluate by workload, serving path, context behavior, license terms, and how reliably it supports the agent or local AI tasks you actually plan to run.

### Why it matters
Open model adoption depends on two things: strong base capability and practical deployment paths. Gemma 4 is important because it pushes both at once. For builders, that means the same project can start with a hosted experiment, move into local testing, and later compare edge or self-hosted deployment without changing the overall model family.

### How it works
Run Gemma 4 on a fixed prompt set from your own workflow. Compare quality, latency, context handling, retry behavior, deployment path, and license fit against nearby open models before adopting it.


## Use Cases
- Local assistant prototypes: Gemma 4 is useful when you want to test an assistant experience on local machines or controlled infrastructure before relying on a remote model service.
- Image-aware workflows: Because the family is positioned around multimodal capability, it is a candidate for workflows that combine text prompts with images, screenshots, documents, or visual inspection.
- Agent and tool experiments: Gemma 4 is relevant for teams exploring multi-step workflows, structured responses, and model behavior inside agent-style systems.

## Alternatives
- Use Gemini when you want Google's managed frontier model experience vs Gemini: Gemini is stronger when you want a hosted model with managed product surfaces. Gemma 4 is stronger when open weights, local testing, and deployment control are the main criteria.
- Benchmark against Llama, Qwen, and Mistral before choosing vs other open model families: The right open model depends on your task, hardware, tooling, license expectations, and inference budget. Gemma 4 deserves evaluation, but it should not be selected only because it is new.

### Getting Started
- Start with the Google DeepMind overview: https://deepmind.google/models/gemma/gemma-4/
- Review official launch details: https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/
- Inspect the Hugging Face models: https://huggingface.co/collections/google/gemma-4

### FAQ
- What should I check before using Gemma 4?
  - Run Gemma 4 on a fixed prompt set from your own workflow. Compare quality, latency, context handling, retry behavior, deployment path, and license fit against nearby open models before adopting it.
- Is Gemma 4 open source?
  - Gemma 4 is listed with Apache-2.0 based on the official source links in this profile. Re-check the repository, model card, or docs before production use.
- Who should evaluate Gemma 4?
  - Gemma 4 is most worth evaluating for developers evaluating open models for local or self-hosted AI apps.
- Is Gemma 4 a replacement for Gemini?
  - No. Gemma 4 is better understood as Google's open model family for developers who need more control, while Gemini remains Google's hosted flagship model line.
- Should I use Gemma 4 for agent workflows?
  - It is worth testing for agent-style workflows, especially where local control or open weights matter, but you should benchmark structured output, tool behavior, latency, and failure modes on your own tasks.
## Why It Matters
Gemma 4 matters because it moves Google's open model line closer to practical agent and on-device use. The family spans smaller edge-oriented models and larger workstation-class models, making it useful for teams that want to test local inference, multimodal understanding, and structured tool workflows without starting from a closed hosted model.


## Best For
- Developers evaluating open models for local or self-hosted AI apps
- Teams that need multimodal reasoning without depending only on hosted APIs
- Builders comparing small edge models against larger workstation-grade open models
- Researchers and product teams that want Apache 2.0 model weights from a major lab

## Not For
- Users who want a fully managed consumer assistant
- Teams that do not want to handle model serving, safety testing, or deployment details
- Workflows that require guaranteed hosted SLA support from the model provider

## What It Actually Does
- Open model family with practical size range: Gemma 4 includes multiple model sizes, from edge-oriented variants to larger models for more demanding workloads.
  - Why it matters: That range lets builders choose between local responsiveness, hardware cost, and model capability instead of treating open AI as one deployment pattern.
- Multimodal and agentic workflow focus: Google positions Gemma 4 for more than simple chat, including multimodal inputs and structured workflows where models need to reason across steps.
  - Why it matters: This makes it more relevant to builders working on assistants, tool-using agents, document workflows, and visual understanding tasks.
- Broad ecosystem support: The launch connects Gemma 4 to Google AI Studio, AI Edge Gallery, Hugging Face, and common local inference tools.
  - Why it matters: A model family is easier to evaluate when developers can try it through familiar runtimes rather than waiting for a single official serving path.

## Typical Use Cases
- Local AI product experiments: Use Gemma 4 to test whether a feature can run on local hardware or self-hosted infrastructure before committing to a hosted-only architecture.
- Multimodal document and image workflows: The family is relevant for apps that combine text with image understanding, such as visual review, document triage, and assistant-style analysis.
- Agentic tool workflows: Gemma 4 is worth evaluating when a workflow needs reasoning across steps, structured outputs, or tool-oriented behavior.
- Edge and mobile AI prototypes: The smaller variants are aimed at low-latency and edge use cases where sending every request to a remote model is not ideal.

## How It Compares
- Choose Gemma 4 when open weights and local control matter vs Gemini: Gemini remains Google's flagship hosted model stack, while Gemma 4 is the better fit when you want downloadable weights, local experimentation, and more deployment control.
- Compare carefully against other open model families vs Llama, Qwen, Mistral: Gemma 4's appeal is its Google research lineage, Apache 2.0 license, and multimodal direction, but teams should still benchmark it against nearby open models on their own tasks.

## Command Line
### Run a workstation model with Ollama
Use this after installing Ollama and confirming your machine has enough memory for the 26B variant.

```bash
ollama run gemma4:26b
```
### Run the smaller edge-oriented variant
Use the smaller variant when you want a lighter local test before trying larger Gemma 4 models.

```bash
ollama run gemma4:e2b
```

## Facts
- Category: models
- Resource type: model
- Open source: yes
- License: Apache-2.0
- Last verified: 2026-04-18

## Capabilities
- local-inference

## Structured Use Case Tags
- local-ai
- self-hosted-ai

## Getting Started
- Read the Google DeepMind overview: https://deepmind.google/models/gemma/gemma-4/
- Read the launch post: https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/
- Open the Hugging Face collection: https://huggingface.co/collections/google/gemma-4
- Try compatible models in Google AI Studio: https://aistudio.google.com/
- Open the Ollama library page: https://ollama.com/library/gemma4

## Links
- Homepage: https://deepmind.google/models/gemma/gemma-4/
- Docs: https://ai.google.dev/gemma
- Demo: https://aistudio.google.com/
- Source: https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/
- Source: https://developers.googleblog.com/bring-state-of-the-art-agentic-skills-to-the-edge-with-gemma-4/
- Source: https://huggingface.co/collections/google/gemma-4
- Source: https://ollama.com/library/gemma4

## Structured Outputs
- JSON: https://www.openagent.bot/models/gemma-4.json
- Markdown: https://www.openagent.bot/models/gemma-4.md
- Canonical: https://www.openagent.bot/models/gemma-4