PomegraArena, the San Francisco AI evaluation platform, raised a $200M Series B at a $3.1B valuation led by Lightspeed and Khosla, nearly doubling its January value.
Key Takeaways
- Arena raised a $200M Series B at a $3.1B valuation, announced October 8, 2026.
- Lightspeed Venture Partners and Khosla Ventures co-led. Annualized revenue now exceeds $100M.
- The round follows a $150M Series A at $1.7B in January 2026.
Lead
Arena, the AI model-evaluation company formerly known as LMArena, announced a $200 million Series B on October 8, 2026, valuing the business at $3.1 billion. Lightspeed Venture Partners and Khosla Ventures co-led the round. The company said annualized revenue has passed $100 million, a figure it reached roughly eight months after closing its previous round.
Who Else Invested in Arena's Series B?
Salesforce Ventures, 01 Advisors, Dell Technologies Capital, Acrew Capital and Endeavor Catalyst joined as new investors. Existing backers a16z, Felicis, AMP PBC, QuantumLight and The House Fund also took part. The company did not disclose how the $200 million splits between new and returning money.
The investor list mixes financial firms with two corporate strategic arms, Salesforce and Dell. Both sell AI products and infrastructure to enterprises, which are the buyers most likely to want a third-party read on model quality.
How Fast Has the Valuation Moved?
Arena's valuation rose from $1.7 billion in January 2026 to $3.1 billion in October, an increase of about 82% in nine months. The January Series A raised $150 million, when the company still operated as LMArena.
Against more than $100 million in annualized revenue, the new price works out to about 31 times run-rate sales. That multiple is steep, though not unusual for AI infrastructure companies growing at this pace. The round implies investors are pricing Arena as a durable piece of AI infrastructure rather than a popular leaderboard.
What Does Arena Actually Do?
Arena runs a public platform where users compare anonymous model outputs side by side and vote on the better one. The approach grew out of the Chatbot Arena project at UC Berkeley's Sky Computing Lab, and the aggregated votes now feed rankings that model developers cite when they release new systems. CEO Anastasios Angelopoulos leads the company.
The company reports 350 million sessions across its platform and 62 million votes spanning text, image and other modalities. Its Agent Arena product, launched less than five months ago, has logged 7 million sessions. Arena says tens of millions of people from more than 150 countries visit each month, and it has tracked more than 1,000 model evaluations.
What Is the Arena Alignment Index?
The Arena Alignment Index is a preview measurement framework, released alongside the funding, that tests whether frontier models and agents act consistently with human intent. It scores three behaviors. Unauthorized Action captures a model doing more than the user asked. False Attribution covers a model crediting statements to the user that the user never made. Deceptive Completion tracks a model claiming a task is finished when it is not.
Arena evaluated more than 20 frontier models and open-sourced 375,000 data points behind the first results. The index moves the company from ranking which answer people prefer to measuring whether a model can be trusted to act on its own. That is a harder claim to defend, and a more valuable one if it holds.
What Are the Open Questions?
Crowdsourced voting has known weaknesses. Preference rankings can be gamed where safeguards are thin, and voters may favor answers that sound correct over answers that are correct. AI agents have also shown they can exploit weaknesses in evaluation systems, which matters for a company now scoring agent behavior.
Arena's neutrality is the second question. A ranking body gains authority from independence, and a $3.1 billion company with strategic investors from the enterprise software and hardware industries will face scrutiny on that point. Competition is a third factor, with smaller rivals such as Barcelona-based Galtea, which raised $3.2 million in March 2026, targeting the same evaluation demand from Europe.
Outlook
Arena has roughly doubled its valuation in nine months on the back of a $100 million revenue run rate and a widening product set that now includes agents and alignment testing. The proceeds are likely to fund the Alignment Index and Agent Arena, though the company has not published a detailed breakdown of spending. The next test is whether labs and enterprise buyers treat the Alignment Index as a reference standard, and whether Arena can keep its rankings credible as its investor base and commercial ties grow.