Measure and Advance the Frontier of AI for Real-World Use

Arena (formerly LMArena) is a community-powered platform for understanding AI performance in the real world. Created by researchers from UC Berkeley, Arena lets tens of millions of builders, researchers, and creative professionals put frontier AI models to the test, head-to-head, and vote on which one actually performs better.Arena's tech-enabled platform anonymizes model outputs in real time, collects votes at scale, and turns the results into a public leaderboard that the entire AI industry watches. OpenAI, Google, Anthropic, Meta, and DeepSeek all submit models to be ranked, both before they release them publicly and after, to monitor how they hold up in the wild.Arena is headquartered in San Francisco, and is one of the fastest-growing companies in the AI evaluation sector. Arena's leaderboard is built from more than 82 million votes across over 700 million conversations, reaching more than 10 million monthly visitors from over 150 countries.

Arena in Three:

• What problem is Arena solving?

  Arena is solving the challenge of knowing which AI model actually works best for a given task, replacing static, corporate-led benchmarks with real human feedback at scale.

• What's innovative about Arena?

  Arena anonymizes model responses and lets real users vote on quality, generating a continuously-updated, statistically rigorous leaderboard grounded in actual usage rather than curated test sets.

• Who can use Arena?

  Developers, researchers, AI labs, and everyday users who want to compare AI models or contribute their preferences to a public, trusted leaderboard can use Arena.