Analysis
Arena, the crowdsourced AI model leaderboard formerly known as LMArena, has raised a $200 million Series B co-led by Lightspeed Venture Partners and Khosla Ventures, with Salesforce Ventures, Dell Technologies Capital, Endeavor Catalyst, a16z and Felicis also participating, according to TechCrunch.
The new round values Arena at $3.1 billion -- nearly double the $1.7 billion mark it set just ten months ago in a January Series A.
From research project to AI's report card
Arena started in 2023 as a UC Berkeley research project that let users compare AI model outputs head-to-head and vote on which was better. That free, crowdsourced product still draws tens of millions of monthly visitors and has become a reference point model labs cite in their own launch posts. The business model changed in September 2025, when Arena launched AI Evaluations, a paid product that gives labs and enterprises structured performance analytics instead of just leaderboard rankings. Annualized revenue went from $30 million at the time of the January round to $100 million by June -- a more-than-tripling in five months that explains why investors were willing to pay up.
The new product: grading AI on safety, not just quality
The round comes alongside a new alignment category on Arena's leaderboard, ranking models on issues like unauthorized action, false attribution and what the company calls deceptive completion. Arena's own framing is blunt, saying AI is advancing faster than our ability to evaluate it, and that the world needs a neutral third party to measure how safe and aligned AI actually is. OpenAI's models currently top that preliminary alignment leaderboard. Arena previously crossed $100M in annualized revenue on the back of the quality-ranking business alone; the alignment leaderboard is a bet that safety grading becomes just as commercially important.
Competitive and strategic context
Arena's bet is that evaluation becomes a durable layer above the model race, similar to how ratings agencies sit above the companies they grade. That is a different wager than the one most of its backers are also making elsewhere: a16z, Khosla and Salesforce Ventures are all active investors in the foundation-model labs Arena now proposes to grade on safety, an overlap that raises a real conflict-of-interest risk the company has not publicly addressed. Arena's leaderboard credibility depends on being seen as neutral by the same labs that are also, in some cases, its investors' portfolio companies.
The numbers in context
A $3.1 billion valuation on roughly $100 million of annualized revenue is a 31x revenue multiple -- rich even by 2026's AI standards, though in line with other infrastructure-layer AI startups that investors are pricing as category-definers rather than as SaaS businesses. The company has not disclosed profitability or burn.
What to watch
Whether enterprises and labs actually pay for alignment scores the way they pay for quality rankings is unproven -- the alignment leaderboard is new, and Arena's revenue history so far is built entirely on the quality-ranking product. If labs start disputing or gaming their safety scores the way some already contest quality rankings, Arena's claim to neutrality will be tested in public.