Popular AI leaderboard Arena’s valuation nearly doubles to $3.1B in 10 months

Meta description: Arena raised $200 million in a Series B at a $3.1 billion valuation, reflecting rising demand for crowdsourced AI model evaluation and enterprise adoption.

Tags: Arena Series B, AI evaluation, model benchmarking, crowdsourced data, enterprise AI

Featured image: Abstract technology style network of glowing evaluation nodes and voting paths, no people, no logos, no text.

Arena Series B funding has turned a UC Berkeley research project into a $3.1 billion AI evaluation platform. The company said on Thursday it raised $200 million in a round led by Lightspeed Venture Partners and Khosla Ventures, with Salesforce Ventures, 01 Advisors, Dell Technologies Capital, Endeavor Catalyst, a16z, Felicis and others joining. The new valuation nearly doubled from its January Series A in about 10 months, and it follows the company reporting $100 million in annualized run-rate revenue in June.

From a 2023 Research Project to a Free Evaluation Arena

Arena began in 2023 as a research project at UC Berkeley that crowdsourced rankings of AI models. Its consumer platform remains free for users to access. People enter prompts or request vibe-coded projects, then rate which model does it better. Arena claims the platform has tens of millions of monthly visitors. That traffic is the core of its method: instead of relying only on fixed test sets, the rankings come from repeated comparisons made by users across many prompts and tasks.

Product, Revenue and the Benchmark Problem

The commercial shift came in September of last year, when Arena introduced AI Evaluations, a service that provides model labs and enterprises with detailed performance analytics based on community feedback. The timing proved impeccable. This year, AI labs realized that their models were gaming benchmarking tests, finding ways to rack up good scores without truly earning them. Enterprises wanted help determining which model works best for internal needs rather than relying only on standardized benchmarks.

In its funding announcement, Arena said, “AI is advancing faster than our ability to evaluate it, and static benchmarks break down once models recognize they’re being tested.” It added, “The world needs a neutral third party to measure how safe and aligned AI actually is once it’s in the hands of real people. Arena is stepping into that role today.”

Alignment Scores and the Next Test

Arena has now added a new category to its leaderboard: alignment, a broad term for whether models follow user intent and avoid deceptive or unauthorized behavior. The category ranks models based on unauthorized action (taking actions it wasn’t asked to take), false attribution (wrongly crediting statements or facts to wrong source), and deceptive completion (lying about completing tasks it didn’t do). Currently, a slate of OpenAI’s models is at the top of its preliminary alignment leaderboard, with Claude Opus 5.5 and Claude Fable in sixth and ninth place, respectively.

The $200 million Series B gives Arena more room to expand AI Evaluations as labs and enterprises search for signals that go beyond fixed benchmark scores. The company’s January Series A was $150 million at a $1.7 billion post-money valuation, the company’s value after the new investment, when its annualized revenue was $30 million. By June, it said annualized run-rate revenue had reached $100 million, a step that helps explain the higher valuation.

The current stage is a shift from a research ranking site to a larger AI evaluation business. What matters next is whether the platform’s crowdsourced comparisons can keep producing enough fresh data, and whether its alignment rankings become a reference point for model builders, buyers and researchers as AI systems move deeper into daily use.