Arena, the AI model evaluation company formerly known as LMArena, has raised $200m in a Series B round at a $3.1bn valuation. Lightspeed Venture Partners and Khosla Ventures co-led the round, the company announced on 8 October 2026.
Salesforce Ventures, 01 Advisors, Dell Technologies Capital, Acrew Capital and Endeavor Catalyst joined as new investors. Existing backers Andreessen Horowitz, Felicis, AMP PBC, QuantumLight and The House Fund also took part.
The new valuation is about 1.8 times the $1.7bn post-money figure Arena reached with its $150m Series A in January 2026, TechCrunch reported.
What Arena does
Arena began in 2023 as a UC Berkeley research project, previously known as lmsys.org, that crowdsourced rankings of AI models. On its free platform, users type a prompt, compare answers from different models and vote for the better one. The votes feed public leaderboards that labs and users watch closely.
The company says its platform has hosted 350 million sessions and gathered 62 million votes across text, vision, code, search, video and image tasks, with tens of millions of monthly visitors from more than 150 countries. It has run over 1,000 new model evaluations and released 375,000 data points on Hugging Face.
Its paid business, AI Evaluations, launched in September 2025 and sells detailed analytics drawn from community feedback to model developers and enterprises.
From $30m to $100m in revenue
Arena said it has passed $100m in annualised revenue. At the time of its Series A in January its annualised revenue was about $30m, according to TechCrunch, and the company said in June that it had reached a $100m run rate. These are company-stated run-rate figures rather than reported annual revenue.
Newer products are adding usage. Agent Arena, which tests AI agents, logged 7 million sessions in less than five months after launch, the company said.
A push into alignment testing
With the funding, Arena introduced the Arena Alignment Index, which scores models on three signals: unauthorised action, false attribution and what it calls deceptive completion. The company said its first results cover more than 20 frontier models, and TechCrunch reported that OpenAI models held the top places on the preliminary leaderboard.
Arena argued that fixed benchmarks lose value once models can tell they are being tested, and said it wants to act as a neutral third party measuring how safe and aligned AI systems are.
The funding so far
Arena raised a $100m seed round in May 2025 led by Andreessen Horowitz and UC Investments, the University of California's investment office, at a $600m valuation reported by Bloomberg, according to TechCrunch. Felicis and UC Investments co-led the $150m Series A in January 2026, Reuters reported. Together, the three disclosed rounds add up to $450m.
The company did not say how it will split the new money, saying only that it will be used to measure and advance frontier AI for real-world use. It did not disclose its headcount.
Company website: arena.ai