Arena, the AI leaderboard utilized by everyone, has now reached a valuation of 100 million dollars.

      TL;DR Arena, the AI leaderboard that originated at UC Berkeley, achieved 100 million dollars in annualized revenue just eight months after launching its paid evaluation service.

      Arena, a crowdsourced AI leaderboard that began as a research initiative at UC Berkeley in 2023, has attained 100 million dollars in annualized revenue following the launch of its first commercial product eight months ago. The platform is recognized for allowing users to compare two anonymous AI model responses side by side and vote on which one they think is superior. To date, over 10 million evaluations have been submitted.

      This revenue stems from AI Evaluations, a paid service launched in September that offers model labs and businesses in-depth performance analytics sourced from its user community. By December, this service generated 30 million dollars in annualized revenue, and it has more than tripled since.

      However, there's a caveat regarding the reported figure. Although Arena labels it as ARR, CEO Anastasios Angelopoulos informed TechCrunch that customers pay based on usage, which means the revenue is not recurring in the conventional SaaS model. “Many people don’t realize that our business is profitable at all; they still perceive us as an open-source project,” he stated.

      The heart of EU tech The latest updates from the EU tech landscape, a story from our esteemed founder Boris, and some dubious AI art. It's free, every week, delivered to your inbox. Sign up now!

      Arena currently has no direct competitors remaining. The other crowdsourced AI model-selection startup shut down in March after raising 33 million dollars from a16z crypto’s Chris Dixon. Angelopoulos noted that Arena competes “for the same dollar” as human labeling firms such as Mercor, Surge, and Scale AI, all of which assist model developers in refining their AI post-training.

      This market is expanding rapidly. Handshake’s annualized revenue from AI training nearly doubled from 550 million dollars in January to almost one billion dollars by April, as reported by The Information. Mercor’s annualized revenue also surpassed one billion dollars earlier this year, although a supply chain breach has since strained its relations with key clients like Meta.

      Arena was co-founded by Angelopoulos and Wei-Lin Chiang, both postdoctoral researchers at UC Berkeley, along with Ion Stoica, a UC Berkeley professor and co-founder of Databricks, who mentored the project before its incorporation in April 2025. The company secured 150 million dollars in a Series A round in January at a valuation approaching two billion dollars, raising its total funding to 250 million dollars from investors including Felicis, Andreessen Horowitz, Kleiner Perkins, and Lightspeed.

      The platform now assesses AI models across various domains, including text, coding, vision, image generation, and intricate agent workflows through a newly launched Agent Mode. Its leaderboard has emerged as the standard scorecard for cutting-edge AI models, with labs from OpenAI to Anthropic to Google regularly referencing Arena rankings in their respective announcements. Transforming that influence into a 100 million dollar business within a year indicates that evaluating AI could prove to be nearly as profitable as creating it.