Arena's CEO: Why AI won't crown a single winner
Key points
Key takeaways from the Grit interview with Arena co-founder and CEO Anastasios Angelopoulos (September-2026):
The leaderboard keeps growing. Arena now ranks more than five hundred models. It began with eight models at launch. Roughly ten new releases arrive each week. He expects that cadence to persist.
Knowledge workers dominate. Arena serves tens of millions of users. About twenty-eight per cent are software engineers. Scientists and finance professionals make up much of the rest. These are the workflows Arena wants to measure.
Evaluations-powered revenue. Labs hand Arena about fifty checkpoints each week. Arena scores them against real-usage data on front-end and tool-calling work. It returns an insights dashboard. This product scaled the company past $100 million.
Run-rate milestone. In June-2026, Arena's annualised run-rate revenue hit $100 million. Its valuation stood at $1.7 billion. Growth was driven by its evaluations business.
API lock-in is overrated. Arena calls booming frontier API revenue easy come easy go. If an open-source model offers ninety per cent of the capability at ten per cent of the cost, customers will switch. Arena bets against sticky API moats.
No single winner. People ask weekly whether models will consolidate. Angelopoulos argues the opposite. AI transforms every industry, so many flowers bloom rather than one model dominating.
Utility is measured by giving utility. Every Arena interaction counts as implicit feedback. Query reformulations and downloads are all signals. Users never have to rate models explicitly.
Academia vs founding. Angelopoulos deferred a Stanford professorship to run Arena full-time. He says universities need flexible retention as top AI faculty increasingly leave to found companies.
Read more: Grit
