Arena raises $150M Series A at $1.7B for crowd-sourced model evaluation

Felicis and UC Investments led the round, with a16z, Kleiner Perkins, Lightspeed, The House Fund, LDVP and Laude Ventures taking part. A Felicis general partner and UC Investments' CIO joined the board. The valuation was nearly triple the May seed round's. The platform reaches more than 5 million monthly users in 150 countries and handles 60 million conversations a month. The revenue comes from AI Evaluations. Customers pay by the volume of battles, votes and prompts they consume, not by seat. Labs can test pre-release checkpoints, and enterprises can benchmark models on their own use cases. Momentum has continued. On October 8, Arena announced a $200M Series B co-led by Lightspeed and Khosla Ventures at a $3.1B valuation. Annualized revenue reached $100M by June. For founders, a leaderboard rank now comes from a company paid by the labs it ranks. Researchers led by Cohere Labs' Sara Hooker have accused Arena of favoring large providers, citing Meta's private testing of 27 Llama 4 variants. Treat Arena standings as one signal and keep your own task-specific evals. Third-party measurement is also becoming a funded category, with enterprise buyers as a growing customer base.