Born in 2023 as a research project at the University of California, Berkeley, AI leaderboard provider Arena has reached $100 million in annual run-rate revenue just eight months after commercial launch.
Arena is best known for its popular crowdsourced AI model performance leaderboards generated from over 10 million user ratings. The consumer website allows users to enter prompts to send to the two models. The user then selects which model gave better results.
Arena’s popular AI model leaderboards are free and available to the public, but the company began monetizing its platform in September with the introduction of AI Evaluation, a service that provides detailed performance analysis collected from the community to model labs and enterprises.
Arena’s rapid revenue growth shows that its commercial services are popular not only with customers but also with the rater community. The evaluator community is often drawn to this platform for early access to the latest (often unreleased) AI models.
“A lot of people don’t even realize that our business makes any money at all. People still look at us like an open source project,” Arena co-founder and CEO Anastasios Angelopoulos told TechCrunch.
Arena refers to its revenue milestones as ARR, a term traditionally used to describe annualized recurring revenue, but Angelopoulos clarified that the company charges customers for “consumption,” meaning the revenue is not recurring.
Although Arena has no direct competitors, another crowdsourced AI model selection startup, Yupp, shut down in March, Angelopoulos said the company competes “on a dollar-for-dollar basis” with human labeling startups like Mercor, Surge, and Scale AI. Both of these startups help modelers improve their AI after training.
As AI providers strive to maximize model performance, the demand for post-training refinement services continues to grow. When Arena announced in January that it had raised $150 million in Series A funding at a post-money valuation of $1.7 billion, the company had annual revenue of $30 million.
Elsewhere, Handshake’s total annual revenue from AI training has nearly doubled since January, from $550 million to nearly $1 billion, The Information reported in April. According to the information, Melkor’s annual sales also exceeded $1 billion earlier this year, up from $500 million in September last year.
Arena ranks models based on a variety of tasks such as text, coding, vision, image generation, and complex, long-running workflows with the recently introduced agent mode.
Along with Angelopoulos, Arena was co-founded by Wei-Lin Chiang, also a postdoctoral fellow at the University of California, Berkeley, who serves as the startup’s CTO. The startup was also co-founded by Ion Stoica, a distinguished professor at the University of California at Berkeley and co-founder of Databricks, who advised the project before it was 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.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
