Arena raises $200M at $3.1B valuation and previews AI safety index

Arena, the platform where users vote on which AI model produces the better answer, has raised $200 million in a Series B round that values the company at $3.1 billion. According to a report on dev.to citing TechCrunch, the round was co-led by Lightspeed Venture Partners and Khosla Ventures and was announced on October 8, 2026. The valuation nearly doubles what investors paid roughly ten months earlier.

The round

Arena announced the deal in a blog post of its own. New investors joining the Series B include Salesforce Ventures, 01 Advisors, Dell Technologies Capital and Endeavor Catalyst. Existing backers a16z, Felicis, AMP PBC, QuantumLight and The House Fund also participated, per Arena's announcement.

The company's valuation has climbed quickly. Its Series A, a $150 million round led by Felicis and UC Investments, closed on January 6, 2026 at a $1.7 billion post-money valuation. TechCrunch had reported that figure in June. Arena has not detailed how it plans to deploy the new capital beyond continued product work and hiring.

How Arena makes money

The company began in 2023 as a UC Berkeley research project that crowdsourced rankings of AI models, and incorporated as a business in April 2025, as TechCrunch previously reported. Its co-founders are chief executive Anastasios Angelopoulos and chief technology officer Wei-Lin Chiang. UC Berkeley professor Ion Stoica, also a Databricks co-founder, advised the project before incorporation.

The public leaderboard itself is free: a user submits a prompt, two models respond, and the user picks the better reply. Arena says the platform has recorded 350 million sessions and 62 million votes from participants in more than 150 countries.

Revenue comes from AI Evaluations, a paid product launched in September 2025 that provides model makers and enterprises with detailed analysis. Arena now says it has passed $100 million in annualized revenue, up from $30 million at the time of its January round, according to TechCrunch. Angelopoulos has described the pricing as consumption-based rather than recurring subscription revenue.

Arena told TechCrunch in June that it has no direct competitors, though it competes "for the same dollar" with data-focused firms such as Mercor, Surge and Scale AI.

A new index for unsafe behavior

Alongside the funding news, Arena released a preview of its Arena Alignment Index, which scores models on three signals: Unauthorized Action, False Attribution and Deceptive Completion. These appear to cover a model acting without permission, crediting the wrong source, and claiming work it did not finish. Arena says the definitions draw on the system cards that OpenAI and Anthropic publish to document model risks.

In the preliminary results, OpenAI models sit at the top of the index, TechCrunch reports, while Claude Opus 5.5 ranks sixth and Claude Fable ninth. Arena argues the industry needs this kind of measurement, writing that "AI is advancing faster than our ability to evaluate it," and positioning itself as "a neutral third party to measure how safe and aligned AI actually is."

Why it matters

The round is a strong signal of investor confidence in AI evaluation as a business category, not just a research exercise. Arena has turned crowdsourced preference data into a nine-figure-revenue operation in about a year as a company, and its valuation has nearly doubled in ten months.

The Alignment Index also matters for developers. Anyone building agentic products, where models act autonomously, faces exactly the failure modes the index measures, such as unauthorized actions and false claims that a task was completed. A public score for these behaviors is useful even in preview form.

There are caveats to keep in mind. Arena's rankings reflect crowd preferences, so they should be treated as one input among several when choosing a model. Its paid evaluation service is sold to the very AI labs whose models it ranks, which is worth remembering when reading the results. The real test of the index's influence will be whether model makers begin citing it in their own launch announcements.