Arena, AI Leader, Reaches $100 Million in Revenue

Arena, a meteoric rise in the world of AI

In just eight months since the launch of its commercial service, Arena, a company specializing in performance rankings of artificial intelligence models, has achieved an impressive annualized revenue of $100 million. This rapid success is all the more remarkable given that the company started as a research project at the University of California, Berkeley, in 2023.

Arena has made a name for itself with its leaderboard system, which evaluates the performance of AI models based on contributions from over 10 million users. The process is simple yet effective: a user submits a query to two AI models and then chooses the one that responded best. This participatory approach has allowed Arena to stand out in the industry.

A rapidly expanding business model

Although access to Arena's leaderboard is free for the general public, the company began monetizing its platform last September. It now offers AI Evaluations, a paid service that provides laboratories and businesses with detailed analyses of model performance based on evaluations from its vast community. This new offering quickly found its audience, contributing to the company's rapid revenue growth.

The popularity of Arena's commercial services reflects the growing interest among businesses for precise and reliable performance analyses. Users of the platform are often attracted by the opportunity to access the latest AI models, which are often still in development.

A perception to change

Anastasios Angelopoulos, co-founder and CEO of Arena, emphasized that despite the commercial success, many still perceive Arena as an open-source project. This perception is partly due to the company's academic origins and its commitment to the developer community.

While Arena uses the term ARR (Annual Recurring Revenue) to describe its revenues, Angelopoulos clarifies that the company charges its clients based on consumption. This means that revenues are not necessarily recurring but depend on the actual usage of the services by clients.

Indirect competition

Arena does not have direct competitors in the current market. Another startup, Yupp, which offered a similar crowdsourced AI model selection service, shut down in March. However, Arena finds itself competing with human labeling companies like Mercor, Surge, and Scale AI. These companies also assist model creators in improving their AIs after the initial training.

The appetite for post-training refinement services continues to grow among AI providers, fueling demand for Arena's offerings. In January, Arena raised $150 million in a Series A round, reaching a post-money valuation of $1.7 billion, while its annualized revenue was $30 million.

A booming market

Meanwhile, other players in the market are also experiencing rapid growth. Handshake's annualized revenue, for example, has nearly doubled since January, rising from $550 million to nearly $1 billion. Similarly, Mercor saw its annual revenue exceed $1 billion earlier this year, compared to $500 million last September.

Arena offers model rankings across various tasks, ranging from text processing to image generation, coding, and computer vision. Its recently introduced Agent mode allows for the evaluation of complex and lengthy workflows, further enhancing the appeal of its platform.

A strong founding team

Arena was co-founded by Anastasios Angelopoulos and Wei-Lin Chiang, both postdoctoral researchers at UC Berkeley. Wei-Lin Chiang serves as the startup's CTO. Ion Stoica, a renowned professor at UC Berkeley and co-founder of Databricks, also played a key role in advising the project before its transformation into a company in April 2025.

So far, Arena has raised a total of $250 million from prestigious investors such as Felicis, Andreessen Horowitz, The House Fund, LDVP, Kleiner Perkins, Lightspeed Venture Partners, Laude Ventures, and UC Investments. This solid funding base underscores investors' confidence in Arena's potential for continued growth.