Design Arena

Overview

Design Arena is the world's largest crowdsourced benchmark for AI-generated design, incubated by Y Combinator in Summer 2025. Using real users' preference votes and a Bradley Terry scoring system (Elo ranking), the platform ranks more than 50 AI models in real time on design aesthetics and functional quality, giving developers and designers an objective reference for choosing AI models.

Key Features

• Head-to-head design comparisons: Shows users designs generated by two AI models from the same prompt and collects real human preference data through crowdsourced voting
• Multi-category evaluation: Covers 13+ design categories including website generation, UI components, image generation, video creation, 3D design, game development, data visualization, and logo design
• Real-time Elo leaderboard: Dynamically updates model rankings from head-to-head results, tracking 50+ LLMs, 12+ image models, 4+ video models, and 22+ audio models
• Micro Evals automated evaluation: Provides automated evaluation for code generation, testing how AI coding tools such as Lovable, Bolt, and v0 perform on identical prompts
• Transparent methodology: Publishes system prompts and evaluation methods, and uses CAPTCHA verification to block bot voting and keep the data authentic

Use Cases

When companies choose AI design tools, product managers can use Design Arena's leaderboards to see how different models actually perform on specific tasks (such as logo generation or website design), saving internal testing time and resources. AI model developers can use the platform to gather performance data based on real user feedback and identify their models' aesthetic and functional strengths and weaknesses.

Unique Advantages

Unlike traditional benchmarks that rely on technical metrics, Design Arena's core difference is "measuring taste" — an evaluation system driven by real user behavior that creates AI model leaderboards reflecting genuine human aesthetic preferences. In its first month online, the platform attracted 47,000+ users from 135 countries and processed tens of thousands of real comparisons.