Chatbot Arena Hits $100M ARR in 8 Months, Valued at $1.7B

新智元 · wechat · 2026-07-05

Chatbot Arena, an LLM evaluation platform created by a UC Berkeley team, reached an annualized revenue of $100 million just 8 months after launching its commercial services, with a valuation surpassing $1.7 billion. In January this year, it closed a $150 million Series A funding round led by Felicis and UC Investments. The platform has accumulated over 10 million user evaluations, 700 million conversations, and 82 million votes, attracting more than 10 million monthly visitors. About 80% of user queries are entirely new questions.

Arena's core commercial product, AIEvaluations, allows model vendors and large enterprises to pay to mobilize the platform's community of tens of millions of users for deep, authentic evaluations. Top players like OpenAI, Google, Anthropic, and Meta have integrated with it; OpenAI even secretly tested on the leaderboard under the codename "summit" ahead of GPT-5's official release. This is a classic "pick-and-shovel" model—the more intense the competition among LLM vendors, the greater the demand for authentic third-party evaluations.

The company was co-founded by CEO Anastasios Angelopoulos (a Berkeley PhD who studied under Michael I. Jordan), CTO Wei-Lin Chiang (creator of the open-source model Vicuna), and Berkeley professor and Databricks co-founder Ion Stoica. Arena has now launched AgentMode, expanding evaluations from chat to multi-round, long-horizon tasks like coding, debugging, and research, while introducing objective metrics such as task completion rate and hallucination rate.

Original post →

More from Companies & People

Companies & People channel →