Arena, the company that grew out of a 2023 UC Berkeley research project ranking AI chatbots by crowd vote, has raised $200 million in a Series B round that values it at $3.1 billion, the company announced Thursday, Oct. 8.
Lightspeed Venture Partners and Khosla Ventures co-led the round. Salesforce Ventures, 01 Advisors, Dell Technologies Capital, Acrew Capital and Endeavor Catalyst joined, along with earlier backers including Andreessen Horowitz, Felicis and Berkeley's The House Fund, according to Arena's announcement.
A valuation that nearly doubled
The new price is a big jump in a short time. In January, the company, then still called LMArena, announced a $150 million Series A at a $1.7 billion post-money valuation, TechCrunch reported. At that point it said its annualized revenue was about $30 million. Arena says that figure has since passed $100 million, a milestone TechCrunch reported the company first reached in June.
The consumer side of Arena is free: people type a prompt, see answers from two models, and vote for the better one. The money comes from a paid evaluation service, launched in September 2025, that gives AI labs and businesses detailed performance data drawn from that community feedback. Arena says its platform has logged 350 million sessions and 62 million votes, and draws tens of millions of visitors a month from more than 150 countries.
Grading behavior, not just smarts
Alongside the funding, Arena released an early version of what it calls an Alignment Index. Rather than asking which model gives the best answer, it looks at how AI agents behave while doing real work for people. It starts with three measures: an agent taking actions the user did not ask for, an agent wrongly attributing a statement or fact to the user, and an agent claiming it finished a task when it did not.
The company argues that fixed benchmark tests stop being reliable once models learn to recognize them, and that an independent referee is needed now that agents write code and take actions people cannot easily check. Arena says it is publishing the methodology and first results for more than 20 frontier models and will add signals over time.
In the preliminary ranking, several OpenAI models sit at the top, TechCrunch reported, with Anthropic's Claude Opus 5.5 in sixth place.
Why it matters locally
Arena is one of the more prominent startups to come out of a Bay Area university lab during the AI boom, and its leaderboard is widely watched by the same Bay Area companies whose models it ranks. The company says it is hiring.