Arena raises $200 million Series B at a $3.1 billion valuation

Arena, the AI evaluation company behind widely used model leaderboards, said it raised a $200 million Series B at a $3.1 billion valuation and launched an Alignment Index that scores how agents behave in real user sessions.

What Arena announced

On Oct. 8, Arena said it closed a $200 million Series B at a $3.1 billion valuation to expand real-world AI evaluation. The company said the round was co-led by Lightspeed Venture Partners and Khosla Ventures, with participation from Salesforce Ventures, 01 Advisors, Dell Technologies Capital, Acrew Capital, and Endeavor Catalyst. Existing backers including a16z, Felicis, AMP PBC, QuantumLight, The House Fund, and Panos Madamopoulos also joined, Arena said.

Arena said it has exceeded $100 million in annualized revenue. Alongside the financing, it introduced the Arena Alignment Index, which it describes as a measure of how frontier models act in line with human intent once they are used outside static benchmarks.

How the Alignment Index works

Arena said the Alignment Index preview starts with three signals that can be checked against an agent trace from real human use: Unauthorized Action, when a model acts beyond what the user asked; False Attribution, when a model credits a statement, intention, or fact to the user that conflicts with user-provided evidence; and Deceptive Completion, when a model tells a user a task is finished when it is not. The company said it is sharing methodology and initial results for more than 20 frontier models and will expand the leaderboard signal by signal.

Arena framed the product as a complement to Agent Arena, its capability leaderboard for agent workflows such as coding, document analysis, and creative writing. It argued that static benchmarks break down once models recognize they are being tested and that buyers need an independent read of safety and trustworthiness in live use.

Scale and independent confirmation

In the same post, Arena said Agent Arena logged 7 million sessions in less than five months, the broader platform logged 350 million sessions, and users cast 62 million votes across text, vision, code, search, video, and image modalities. It also said it draws tens of millions of monthly visitors from more than 150 countries, has run more than 1,000 new model evaluations, and has open-sourced 375,000 data points for research.

TechCrunch independently reported the $200 million Series B and $3.1 billion valuation on Oct. 8. TechCrunch said Lightspeed and Khosla led the round — Arena's own wording is co-led — and named overlapping new and existing investors. TechCrunch also said Arena originated in 2023 as a UC Berkeley research project, previously raised a $150 million Series A in January at a $1.7 billion post-money valuation when annualized revenue was about $30 million, and reached $100 million in annualized run-rate revenue in June.

Why the raise matters

The financing lands as model labs and enterprises look for evaluation that is harder to game than fixed benchmarks. Arena is selling itself as a neutral measurement layer for both capability and alignment, pairing a free consumer ranking surface with a commercial AI Evaluations product that TechCrunch said launched the prior September. For capital-markets readers, the anchored package is a $200 million Series B, a $3.1 billion valuation, co-leads Lightspeed and Khosla, annualized revenue above $100 million, and a same-day Alignment Index preview with three named behavioral signals.

What is still open

Arena did not disclose ownership percentages, a full post-money capitalization table, or customer contracts tied to the Alignment Index. TechCrunch's "led by" language differs from Arena's "co-led" wording, so this story keeps the company formulation for the lead structure and notes the TechCrunch phrasing. Series A size and the June $100 million run-rate timing are attributed to TechCrunch where the Series B post does not restate them.