Arena AI leaderboard hits $100M revenue eight months after launch

Arena's rapid commercial growth

Arena, the widely-used AI model leaderboard that originated as a UC Berkeley research project in 2023, has reached $100 million in annualized run-rate revenue just eight months after launching commercial services, according to TechCrunch.

The platform is best known for its free public leaderboard that ranks AI models based on crowdsourced evaluations. More than 10 million user assessments have contributed to Arena's rankings, which work by presenting users with responses from two anonymous models and asking them to choose the better output.

Arena began monetizing in September 2025 when it introduced AI Evaluations, a paid service that provides model developers and enterprises with detailed performance analytics derived from its evaluation community. The company's annualized revenue has more than tripled since January 2026, when it stood at $30 million following a $150 million Series A funding round at a $1.7 billion valuation.

Why it matters

Arena's revenue trajectory demonstrates the intense demand for post-training refinement services as AI providers compete to maximize model performance. The company's success also validates a crowdsourced approach to AI evaluation at a time when traditional human labeling services are scaling rapidly—competitors like Mercor and Handshake have reported annualized revenues exceeding $1 billion.

Competing in the post-training market

Co-founder and CEO Anastasios Angelopoulos told TechCrunch that many people don't realize Arena operates as a revenue-generating business, still perceiving it as an open-source project. While Arena uses the term "ARR," Angelopoulos clarified the company charges for consumption rather than recurring subscriptions.

Arena doesn't face direct competitors in the crowdsourced leaderboard space—rival startup Yupp shut down in March 2026. However, the company competes for budget with established human labeling providers including Mercor, Surge, and Scale AI, all of which help model makers refine AI systems during post-training phases.

The broader market for AI training services continues expanding rapidly. The Information reported in April that Handshake's gross annualized revenue from AI training nearly doubled from $550 million in January to approximately $1 billion. Mercor's annualized revenue also topped $1 billion earlier this year, up from $500 million in September 2025.

Platform capabilities and leadership

Arena evaluates models across multiple dimensions including text generation, coding, vision, and image creation. The platform recently added Agent Mode to assess complex, long-running workflows.

Angelopoulos co-founded Arena with Wei-Lin Chiang, a fellow UC Berkeley postdoctoral student who serves as CTO, and Ion Stoica, the prominent UC Berkeley professor and Databricks co-founder who advised the project before its incorporation in April 2025.

The startup has raised $250 million total from investors including Felicis, Andreessen Horowitz, The House Fund, LDVP, Kleiner Perkins, Lightspeed Venture Partners, Laude Ventures, and UC Investments.

These details were first reported by Marina Temkin at TechCrunch.