Arena raises $200 million as its AI evaluation business reaches a $3.1 billion valuation

Arena’s AI evaluation business now carries a $3.1 billion valuation, up from $1.7 billion in January. The company announced a $200 million Series B on October 8, 2026, co-led by Lightspeed Venture Partners and Khosla Ventures. The financing backs its mission to measure how AI performs in real-world use.

Salesforce Ventures, 01 Advisors, Dell Technologies Capital and Endeavor Catalyst also participated. Arena named support from existing investors including a16z, Felicis, AMP PBC, QuantumLight and The House Fund. The new financing follows a $150 million Series A announced in January.

Free comparisons, paid analytics

Arena began in 2023 as a UC Berkeley research project that crowdsourced rankings of AI models. Its consumer platform remains free: people submit prompts or request coding projects, compare the models’ work and vote for the better result. Those judgments provide feedback about how people experience different models.

The commercial layer is AI Evaluations, introduced in September 2025. It sells model labs and enterprises detailed performance analytics based on community feedback, TechCrunch reported. That connects Arena’s public comparison service to a business serving organizations that develop AI or need to choose models for their own work.

Arena said its annualized revenue pace reached $100 million in June, compared with $30 million when it announced the Series A. In the new funding announcement, it said annualized revenue had exceeded $100 million. The June figure describes a run rate, rather than revenue collected over a completed year.

Community feedback is the foundation

The funding announcement offers a picture of that feedback engine’s scale. Arena says Agent Arena logged 7 million sessions in less than five months after launch. Across the entire platform, it reports 350 million sessions and 62 million votes spanning text, vision, code, search, video and image.

Arena also says it has tens of millions of monthly visitors from more than 150 countries. It reports more than 1,000 new model evaluations covering open and proprietary systems, and 375,000 data points released for open AI research. These are company-reported measures of participation and evaluation activity.

The company’s stated ambition is to be a neutral third party for evaluating AI in real use. Its own history illustrates why preference alone is insufficient: Arena says it added factuality measurement after finding that the answer people preferred was not always correct. The financing supports that broader evaluation mission.