Google Gemma 4

Google Gemma 4 is Google's most capable open-weight model family — built for on-device and cloud deployment with frontier-level reasoning, 140+ language support, and Apache 2.0 licensing.

Why Vibe Coders Use It

• True open-weight models — Apache 2.0 license means no restrictions on commercial use, fine-tuning, or redistribution
• On-device to cloud — E2B and E4B run on phones and Raspberry Pi; 26B and 31B handle cloud workloads
• Frontier-class reasoning — 31B ranks #3 on Arena AI text leaderboard, outperforming models 20x its size
• Massive context window — 256K tokens on larger variants for processing entire codebases
• 140+ languages — build multilingual apps without separate translation pipelines
• Vision and audio — native multimodal input for building richer applications

Key Specs

Model Variants

The E2B and E4B models activate a small parameter footprint during inference to preserve RAM and battery life. They run completely offline on phones, Raspberry Pi, and NVIDIA Jetson Orin Nano.

Getting Started

Option 1: Via Google AI Studio (Recommended)

Get a free API key from Google AI Studio — no credit card required.

npm install ai @ai-sdk/google

import { google } from '@ai-sdk/google';
import { generateText } from 'ai';

const { text } = await generateText({
  model: google('gemma-4-31b-it'),
  prompt: 'Build a REST API endpoint for user authentication in Express',
});

console.log(text);

Set your environment variable:

GOOGLE_GENERATIVE_AI_API_KEY=your-api-key-here

Option 2: Via Vercel AI Gateway

No API key needed — Vercel hosts Gemma 4 models directly.

import { streamText } from 'ai';

const result = await streamText({
  model: 'google/gemma-4-31b-it',
  prompt: 'Refactor this React component to use server actions',
});

for await (const chunk of result.textStream) {
  process.stdout.write(chunk);
}

Option 3: Local with Ollama

# Download and run locally — no API key, no cloud costs
ollama run gemma4

# Use with the AI SDK
npm install ai ollama-ai-provider

import { ollama } from 'ollama-ai-provider';
import { generateText } from 'ai';

const { text } = await generateText({
  model: ollama('gemma4'),
  prompt: 'How do I set up database migrations with Drizzle ORM?',
});

Option 4: On Google Cloud (Vertex AI)

For production workloads, deploy via Vertex AI Model Garden with managed infrastructure and autoscaling.

npm install ai @ai-sdk/google-vertex

import { vertex } from '@ai-sdk/google-vertex';
import { generateText } from 'ai';

const { text } = await generateText({
  model: vertex('gemma-4-31b-it'),
  prompt: 'Design a database schema for a multi-tenant SaaS application',
});

When to Use Gemma 4 vs. Alternatives

Use Gemma 4 when you need open-weight models you can self-host, fine-tune, or run on-device without API costs. The E2B/E4B variants are unmatched for mobile and edge deployment. Use Gemini when you want Google's strongest proprietary model with 1M+ context. Use Llama for the largest open-source ecosystem and community. Use Claude for the strongest proprietary reasoning and coding.

Why Solo Founders Should Care

Gemma 4 changes the economics of AI features for indie products:

• Zero marginal cost — run E2B/E4B on-device with no API bills
• No vendor lock-in — Apache 2.0 means you own your AI stack
• Privacy by default — on-device inference means user data never leaves the phone
• Ship faster — 256K context handles entire codebases in agentic workflows
• Global from day one — 140+ languages without separate localization infrastructure

Resources

• Google AI Studio — free API access
• Gemma on Hugging Face
• Vertex AI Model Garden
• Ollama (Local Deployment)
• AI SDK Google Provider
• See the full Gemma profile on LLMReference →