Arena's Oct. 8 Series B: $200M at $3.1B Valuation

Arena announced a $200 million Series B at a $3.1 billion valuation on October 8, 2026, roughly 82% above January's level in about 10 months, with Lightspeed Venture Partners and Khosla Ventures co-leading the round, according to TechCrunch's account. The new financing puts a much higher price on the AI evaluator as it expands from model comparisons into assessments of agent behavior.

Then called LMArena, the company set out the earlier $150 million Series A, its $1.7 billion post-money valuation and a December annualized consumption run rate above $30 million in its January financing release. A June company update placed the annualized revenue run rate above $100 million, months before the new funding price was set. Those run-rate milestones show the pace investors were assessing, while leaving the actual revenue earned over a completed year undisclosed.

The valuation increase behind the $200 million round

On the published figures, Arena's valuation rose to about 1.82 times its January level. That is a large step in a short interval, though it falls short of a literal doubling. The $200 million Series B is the amount raised in the latest financing; the $1.4 billion increase is the difference between the two headline company valuations. Treating the two as the same kind of gain would overstate what the new cash itself represents.

The comparison also has a capitalization caveat. The earlier valuation was explicitly post-money, meaning it includes the capital from that financing. The October figure has no explicit pre-money or post-money designation. The roughly 82% rise is therefore a calculation from the published figures, not an independently normalized comparison of equity values on identical terms.

The lead-investor lineup changed with the higher price. Felicis and UC Investments led the January financing; Lightspeed and Khosla led the October round. Lightspeed was already among the earlier investors, so its new lead position marks a deeper role in an existing relationship.

What Arena's revenue rates show

The two revenue thresholds are both expressed as annualized rates, which makes a direct comparison more useful than placing a single month's receipts beside a year's sales. At the stated thresholds, $100 million is about 3.3 times $30 million. Both disclosures say the rates exceeded those amounts, however, so 3.3 times is a comparison of benchmarks rather than Arena's exact growth multiple. The December rate predates the January financing, and the June rate predates the October financing; neither is a same-day revenue figure for the corresponding valuation.

An annualized run rate takes activity over a shorter interval and expresses what it would amount to if maintained for a full year. It can capture a fast-moving business sooner than a trailing-year figure, but it also depends on the chosen measurement period and the pace continuing. Chief executive Anastasios Angelopoulos told TechCrunch in June that "people still see us as an open source project"; he explained that customers pay for consumption, so the revenue is not contractually recurring. That distinction matters when projecting the rate forward.

The public model-comparison site and the paid evaluation service play different roles in the business. Users contribute judgments about model responses, creating signals that can help AI labs and enterprises assess performance on tasks relevant to them. Paid consumption is the closer measure of commercial demand. The disclosed run rates make the valuation step-up easier to understand, but a full-year income statement, margins and customer retention would give investors a firmer view of how durable that demand is.

Investors are backing a wider evaluation business

The Series B comes as Arena extends its evaluations from preferred model answers toward what agents actually do on a user's behalf. Its Alignment Index preview focuses first on unauthorized actions, false attribution and deceptive completion: an agent acting beyond a request, misrepresenting what a user supplied, or claiming to have finished work it did not complete. These are concrete failure modes for systems that write code, analyze material and carry out multistep tasks. They also give the funding a product direction beyond maintaining a public leaderboard.

The Series B announcement names Salesforce Ventures, 01 Advisors, Dell Technologies Capital, Acrew Capital and Endeavor Catalyst as participants, alongside existing backers a16z and Felicis, and reports 7 million Agent Arena sessions in less than five months, 350 million platform sessions, 62 million votes and tens of millions of monthly visitors from more than 150 countries. These are measures of evaluation activity and reach, rather than a count of paying customers; they explain the scale of the feedback pool on which the service draws. The next meaningful evidence for the financing price will be whether expanded agent evaluations translate into sustained customer consumption over a longer reporting period.