Is my API key or prompt data stored on your servers?

No. Everyday Used Apps uses a strict 100% client-side Bring-Your-Own-Key (BYOK) architecture. Your API keys are saved exclusively in your browser's localStorage and requests are dispatched directly from your browser to the model provider (Google Gemini, OpenAI, Anthropic, or your local Ollama/LM Studio instance). We have zero middleman servers, proxies, or telemetry.

Can I test Google offline models like Gemini Nano or Gemma 2?

Yes. You can test Google Gemini Nano directly inside Google Chrome 127+ without an API key or internet connection. You can also run Google Gemma 2 (2B, 9B, 27B) locally via Ollama or LM Studio with full privacy.

How does the Arena prevent local models from freezing my device?

The tool inspects your device hardware via navigator.deviceMemory and displays pre-flight warnings if a selected model (like Gemma 2 9B or 27B) requires more RAM than your machine has. We also recommend single concurrency (OLLAMA_NUM_PARALLEL=1) and provide an instant 1-click 'Stop Generation' abort button with a 45-second watchdog timer.

Is the AI Model Arena completely free to use?

Yes, 100% free. You can use free tier API keys (such as Google Gemini's free tier or Groq) or completely free offline local models without any subscription or payment.

AI Multi-Model Arena & Research Canvas | Side-by-Side LLM Benchmark

How to Get a 100% Free Google Gemini API Key:

No credit card needed. You get 15 free requests every minute (up to 1,000,000 tokens/minute) directly from Google.

Open Google AI Studio

Visit aistudio.google.com/app/apikey and sign in with your regular Google account.

Create Key

Click the blue "Create API Key" button and copy the string starting with AIzaSy...

Paste in AI Vault

Click "Key Settings" above and paste your key. It saves safely inside your own browser!

Input Tokens: 0

Session Tokens: 0

Requests: 0

Active Quotas: Gemini Free: 15 RPM / 1M TPM

Enter Test Prompt: 0 chars • 0 words

Cloud or 100% Offline (Gemini Nano & Google Gemma 2) run directly from your device.

MODEL A:

Free / Offline

Tokens: -- Latency: -- Speed: -- tok/s

Model A response will render here...

MODEL B:

Free / Offline

Tokens: -- Latency: -- Speed: -- tok/s

Model B response will render here...

Frequently Asked Questions & Privacy Guide

The AI Multi-Model Arena allows developers, writers, and researchers to benchmark generative models side-by-side with 100% private in-browser execution.

Is my API key or prompt private?

Yes, 100%. All keys are stored solely in your local browser’s localStorage. API calls travel directly from your browser to Google Gemini, OpenAI, or your local runner. No data touches our servers.

Can I run Google offline models?

Yes. Google Gemini Nano can be executed offline via Chrome Built-in AI without an API key. Google Gemma 2 (2B, 9B, 27B) can be run locally via Ollama or LM Studio completely air-gapped.

How is device freezing prevented?

The tool auto-profiles your device RAM. If you select a heavy model that exceeds your physical RAM, an advisory warning appears. A 45-second watchdog timer and 1-click Stop button prevent UI thread locks.

Are live speed & token metrics accurate?

Yes. The arena tracks total elapsed latency in milliseconds and subword token counts to give you real-time tokens-per-second throughput comparisons between models.