Arena, which began as a research project at the University of California, Berkeley in 2023, has reached a $100 million annualized run-rate in revenue just eight months after launching its commercial service. The company is best known for a crowdsourced AI model leaderboard built from more than 10 million user evaluations.
How the platform works
On Arena’s consumer website a user types a prompt that is sent to two models; the user then selects which model performed better. The public leaderboard remains free to use, but the company began monetizing the platform in September by launching AI Evaluations, a paid product that delivers deep-dive performance analytics to model labs and enterprises based on community evaluations.
Arena’s rapid revenue growth suggests its commercial offerings are resonating with customers as strongly as the platform appeals to its evaluator community, many of whom are attracted by early access to the latest, often unreleased, AI models.
Business model and clarification
Anastasios Angelopoulos, Arena’s co-founder and CEO, noted that many people still perceive the company as an open-source project and do not realize it is generating revenue. While Arena reports the milestone as ARR (a term traditionally meaning annualized recurring revenue), Angelopoulos clarified the company bills customers for "consumption," so its revenue is not recurring in the usual sense.
Competition and market context
Arena does not have direct, like-for-like competitors following the March shutdown of Yupp, another crowdsourced model-selection startup. However, Angelopoulos said Arena competes “for the same dollar” as human-labeling startups such as Mercor, Surge, and Scale AI, which help model developers with post-training refinement.
Demand for post-training refinement services continues to rise as AI providers seek to maximize model performance. When Arena announced in January that it had raised a $150 million Series A at a $1.7 billion post-money valuation, its annualized revenue at that time was $30 million.
Other firms in the space have also shown rapid growth: The Information reported in April that Handshake’s gross annualized revenue from AI training nearly doubled since January, from $550 million to nearly $1 billion. The Information also reported that Mercor’s annualized revenue topped $1 billion earlier this year, up from $500 million last September.
Services, team, and financing
Arena ranks models across tasks including text, coding, vision, and image generation, and it supports complex, long-running workflows through its recently introduced Agent Mode.
The company was co-founded by Anastasios Angelopoulos and Wei-Lin Chiang, a fellow UC Berkeley postdoctoral researcher who serves as the startup’s CTO, with advisory and founding involvement from Ion Stoica, the UC Berkeley professor and Databricks co-founder. Arena incorporated as a company in April 2025.
Arena has raised a total of $250 million from investors including Felicis, Andreessen Horowitz, The House Fund, LDVP, Kleiner Perkins, Lightspeed Venture Partners, Laude Ventures, and UC Investments.
