source: techcrunch ai: arena, the ai leaderboard everyone uses, is now a $100m business
level: business
arena, which started as a uc berkeley research project in 2023, now generates $100 million in annualized run-rate revenue. the company is famous for its crowdsourced ai model leaderboard, built from over 10 million user evaluations. users type a prompt, see responses from two anonymous models, and pick the better one. while the leaderboard remains free, arena began charging for its ai evaluations service in september. this service gives model labs and enterprises detailed performance analytics from the community.
co-founder and ceo anastasios angelopoulos noted that many still view arena as an open-source project, unaware of its commercial success. the revenue is not recurring but based on consumption. arena competes for budgets with human labeling startups like mercor, surge, and scale ai, which help refine models after training. demand for post-training refinement is growing as ai providers push for better performance. in january, arena raised $150 million at a $1.7 billion valuation when its annualized revenue was $30 million.
other ai training startups also report rapid growth. handshake's annualized revenue from ai training nearly doubled to $1 billion since january, and mercor's topped $1 billion earlier this year. arena ranks models on text, coding, vision, image generation, and complex workflows through its new agent mode. the company was co-founded by angelopoulos, cto wei-lin chiang, and uc berkeley professor ion stoica, who also co-founded databricks. arena has raised $250 million from investors including felicis, andreessen horowitz, and kleiner perkins.
why it matters: arena's growth shows strong commercial demand for ai evaluation tools, which help model developers improve performance using real user feedback.
source: techcrunch ai: arena, the ai leaderboard everyone uses, is now a $100m business