Arena says its AI leaderboard business reached $100M run-rate revenue

In short: Arena, the company behind a widely used AI model leaderboard, says it has reached $100 million in annualized run-rate revenue.

What happened

Arena started as a research project at UC Berkeley in 2023 and later became a company. It is best known for a free public website that compares AI models. A user types a prompt, Arena sends it to two models, and the user picks which answer is better (like a taste test, but for writing and coding).

According to Arena, it has collected more than 10 million user evaluations this way. The company now says that, eight months after launching its commercial service, it has reached $100 million in annualized run-rate revenue. Annualized run-rate is a way to estimate yearly revenue based on the recent pace of sales.

Arena began making money in September when it launched a paid product called AI Evaluations. This service sells more detailed performance reports to AI labs and other companies, using data gathered from Arena’s community.

Arena’s CEO, Anastasios Angelopoulos, told TechCrunch that many people still think of Arena as a free, open research project. He also said Arena’s revenue is based on “consumption,” meaning customers pay based on usage, so the revenue is not guaranteed to repeat each month.

Why it matters

As more companies build and sell AI models, they need trustworthy ways to compare quality and spot weaknesses. Arena is turning a popular public scoreboard into a paid service for businesses, which shows there is real demand for independent, large-scale AI testing.

Source: TechCrunch AI