Overview
LMArena is the leading community-driven platform for evaluating AI models. Users enter a prompt that is sent to two anonymous models, then vote on which response is better. Aggregated across millions of votes, this produces Elo-style leaderboards that have become an influential, real-world benchmark for large language models across text, vision, coding and agentic tasks. The platform grew out of the UC Berkeley 'Chatbot Arena' research project.
Funding and Scale
LMArena spun out of Berkeley and raised a $100 million seed led by Andreessen Horowitz and UC Investments at a $600 million valuation, followed by a $150 million Series A in early 2026 valuing the company at $1.7 billion. Co-founded by Anastasios Angelopoulos, Wei-Lin Chiang and Berkeley professor Ion Stoica, it serves millions of monthly users and tens of millions of conversations from around the world.
Why It Matters
Static benchmarks are easy to overfit and quickly go stale. LMArena's human-preference, head-to-head methodology has become a trusted signal that AI labs cite when releasing new models, and its private and enterprise arenas help organizations evaluate models on their own data.