The ultimate client-side LLM battleground.
AI Arena is a powerful, open-source interface for comparing AI models side-by-side. Run prompts across 68+ providers and 1000+ models (including local LLMs) simultaneously to benchmark performance, speed, and quality.
Live Demo • Report Bug • Request Feature
Compare complex web-generation tasks: responsive HTML/CSS layouts, interactive JS, SVG/graphics, and more – all from a single prompt.
Analyze latency and output quality side-by-side.
- 🤯 Massive Model Support: Access over 1000+ models from 68 providers (OpenAI, Anthropic, Google, Grok, Mistral, and many more).
- 🔒 Client-Side Architecture (No Backend): AI Arena is a frontend-only application. Your API keys and prompts are sent directly from your browser to the AI providers. No middleman server, no data logging, enhanced privacy.
- ⚔️ Flexible Comparison:
- Cross-Model: Compare GPT-4 vs. Claude 3.5 vs. Gemini 1.5.
- Cross-Provider: Compare the same model (e.g., Llama 3) hosted on different providers (Groq vs. Fireworks vs. Local) to test latency and token speed.
- 🏠 Local LLM Support: Seamlessly integrate with local inference servers (like Ollama or LM Studio).
- 📊 Real-time Metrics: View generation time and performance stats for every output.
AI Arena is built on top of the following open-source projects:
- Vercel AI SDK: Powers the unified API interface and real-time streaming capabilities.
- models.dev: The comprehensive registry used to source the 68+ providers and 1000+ models available in the app.
You can use the hosted version at arena.jit.dev or run it locally for full control.
- Node.js (v18 or higher)
- pnpm (recommended) or yarn
-
Clone the repository:
git clone [https://github.com/andronov04/aiarena.git](https://github.com/andronov04/aiarena.git) cd aiarena -
Install dependencies:
pnpm install # or yarn install -
Run the development server:
pnpm run dev # or yarn dev -
Open your browser: Navigate to
http://localhost:3000to start comparing models.
- Select Models: Click on the "Add model" button or the model dropdown to search through the extensive list of supported providers.
- Configure Keys: Since the app is client-side, you will need to input your API keys for the respective providers (saved locally in your browser).
- Enter Prompt: Type your prompt in the sidebar. You can test code generation, creative writing, or logic puzzles.
- Compare: Hit generate and watch the models stream their responses in parallel.
We welcome contributions! Whether it's adding a new provider, fixing a bug, or improving the UI.
Distributed under the MIT License. See LICENSE for more information.