Code Arena’s leaderboard now has 489,150 votes, ranking 104 AI models on their ability to build fullstack web applications. The platform evolved from a frontend-only tool to testing complex features like PostgreSQL databases, API integrations, and live deployments.
Launched in 2025 focusing on UI components and frontend code, Code Arena expanded on July 2, 2026. It added support for authentication, Row Level Security, persistent sandboxes with hot reloading, and direct deployment to Vercel. This shift marks a major step towards evaluating real-world coding capabilities beyond prototypes.
The ranking is decided through community votes after users build web apps with competing AI models. Anthropic’s claude-opus-5-max leads with a score of 1725, closely followed by Moonshot’s kimi-k3-max at 1682. Arena.ai’s growth mirrors this momentum, reaching a $100 million annualized run rate within eight months and attracting over 10 million monthly users.
This leaderboard offers a live glimpse into which AI models excel at practical coding. It influences developer tools, cloud infrastructure investments, and the competitive landscape of AI-powered software.
Material is for informational purposes and not financial advice.
