The world's most trusted AI evaluation platform — crowdsourced, real-world model rankings
HQ: San Francisco, CA, United States | Founded: 2023 | Employees: ~100 (2026, est.) | Stage: Late Stage VC (Series A) | Website: https://lmarena.ai
LMArena (formerly LMSYS) began as an academic research project at UC Berkeley's Sky Computing Lab: in May 2023, researchers Anastasios Angelopoulos and Wei-Lin Chiang — PhD students working with professor Ion Stoica — launched Chatbot Arena, a crowdsourced platform where users pit two anonymous AI models against each other and vote on the better response. The Elo-style rankings it produces became the most-cited, most-watched measure of AI model quality in the world, referenced by OpenAI, Google, Anthropic, Meta, and Chinese labs alike.
In April 2025 the project incorporated as a company, LMArena, led by the original team, and raised a $100M seed at a $600M valuation (May 2025, co-led by Andreessen Horowitz and UC Investments). In January 2026 it raised a $150M Series A at a $1.7B valuation (led by Felicis with UC Investments), tripling its valuation in about eight months and bringing total funding to $250M.
The company's product is an evaluation platform that moves beyond static benchmarks: humans judge models in real-world, side-by-side comparisons at massive scale, producing leaderboards, preference data, and evaluation APIs that model developers, enterprises, and researchers use to decide which models to ship. Its annualized consumption run rate surpassed $30 million in December 2025 — less than four months after launching the commercial evaluation product. LMArena is headquartered in San Francisco while maintaining deep ties to UC Berkeley, positioning itself as the neutral, trusted referee for the AI industry.
LMArena owns the most influential brand in AI evaluation: Chatbot Arena, the crowdsourced leaderboard that every frontier lab watches, has effectively become the standard for measuring real-world model quality — humans comparing models side-by-side, blind, at scale. The company's trajectory is remarkable even by AI standards: a $100M seed at $600M (May 2025), a $150M Series A at $1.7B (January 2026), and a $30M+ annualized consumption run rate within four months of product launch. The bull case is that as AI becomes ubiquitous, trusted third-party evaluation becomes critical infrastructure with a natural subscription/consumption business; the bear case is that labs could build their own evaluation, incumbents (OpenAI, Google, scale-up platforms) compete aggressively, and the leaderboard's influence depends on continued neutrality and scientific credibility.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki