Hajra Naz
Arena’s commercial business is taking off.
Eight months after introducing its paid services, the AI startup has hit $100 million in annualized run-rate revenue. The company started as a research project at the University of California, Berkeley, back in 2023. It has since evolved into a household name for benchmarking of AI models.
From Research Project to AI Startup
Arena first gained attention through its public AI leaderboard.
The platform lets users compare two AI models without revealing their names. A user enters a prompt, receives two different responses, and simply chooses the better one.
Hundreds of thousands of people have taken part in the intervening years.
Over 10 million assessments have been submitted by users, according to the firm. Arena’s widely used leaderboard, which many developers and AI companies use to quantify model performance, is powered by those votes.
The Leaderboard Is Free, but the Business Isn’t
Arena’s public leaderboard remains free to use.
Both started making money in September, but the company began its AI evaluations after launch. There’s also a service for labs and other businesses that want more granularity in the performance of their models
Customers get detailed analytics about user behavior based on data collected from Arena’s massive testing crowd rather than just benchmark scores.
Revenue Has Grown Rapidly
Arena stated that demand for its commercial products has increased significantly faster than anticipated.
Since it typically offers early access to new AI models, including those that have not yet been publicly released, a lot of people turn to the platform.
It has also turned that active community into one of the company’s biggest advantages.
“In a lot of ways, people still think that we are just an open-source project,” Anastasios Angelopoulos, co-founder and CEO of Arena, told TechCrunch in an interview. “They think we are just running a business.
Read More: Meta May Launch Its Own Prediction Market, Zuckerberg Says
Revenue Isn’t Subscription-Based
Arena describes its milestone as annualized recurring revenue, or ARR.
However, Angelopoulos explained that customers actually pay based on how much they use the service.
In other words, the company’s revenue comes from usage rather than fixed subscriptions.
Competition Comes From a Different Direction
Arena doesn’t see many direct rivals.
One similar startup, Yupp, shut down earlier this year.
Instead, Arena competes with companies that provide human feedback to improve AI models after training. Those companies include Mercor, Surge, and Scale AI.
Although their products differ, they often compete for the same customer budgets.
Demand for AI Evaluation Is Growing
As AI companies race to build better models, spending on evaluation and post-training services continues to rise.
That trend has helped Arena expand quickly.
When the company announced its Series A funding round in January, its annualized revenue stood at $30 million. It has more than tripled that figure in just a few months.
Other companies in the same market are also growing rapidly.
Handshake’s AI training business reportedly increased its annualized revenue from $550 million to nearly $1 billion earlier this year. Mercor also crossed the $1 billion mark after doubling its revenue over the past several months.
Arena Tests More Than Chatbots
Arena evaluates AI models across several categories.
These include text generation, coding, image creation, and vision tasks.
The company has also introduced Agent Mode, which measures how well AI systems perform longer, more complex workflows.
Read More: Nobody Has Figured Out How AI Companies Should work with the government
Built by UC Berkeley Researchers
Researchers from UC Berkeley founded Arena. The company is led by CEO Anastasios Angelopoulos.
The startup’s chief technology officer is Wei-Lin Chiang, a former postdoctoral researcher at UC Berkeley.
Ion Stoica, UC Berkeley professor and co-founder of Databricks, also completes the founding team. Before Arena was an actual company in April 2025, he had already been helping forge the idea.
Backed by Major Venture Capital Firms
Arena has raised $250 million in funding so far.
Its investors include Felicis, Andreessen Horowitz, The House Fund, LDVP, Kleiner Perkins, Lightspeed Venture Partners, Laude Ventures, and UC Investments.
The fresh funding gives Arena additional resources as demand for AI evaluation tools continues to grow across the industry.
