[State of Evals] LMArena's $1.7B Vision - Anastasios Angelopoulos, LMArena - Latent Space: The AI Engineer Podcast Recap

Podcast: Latent Space: The AI Engineer Podcast

Published: 2026-01-06

Duration: 24 min

Guests: Anastasios Angelopoulos

Summary

Anastasios Angelopoulos discusses LMArena's evolution, the company's funding and scaling strategies, and the importance of real-world AI evaluations. He addresses recent critiques while outlining future goals for the platform.

What Happened

Anastasios Angelopoulos from LMArena discusses the company's journey from an academic project at Berkeley to a significant player in AI evaluation. Initially incubated by Ansh, LMArena transitioned into a company to scale their platform, which measures AI capabilities based on real-world usage and organic feedback.

Angelopoulos highlights the importance of starting a business to provide the necessary resources for scaling their platform. With $100 million raised, the funds are primarily allocated towards platform operations and hiring, emphasizing responsible spending to ensure longevity and adaptability.

The platform boasts more than 5 million monthly active users and facilitates around 250 million conversations, making it a major consumer platform for large language models (LLMs). Despite not reaching the scale of ChatGPT, LMArena engages a diverse audience, with a significant portion involved in software development.

Angelopoulos addresses the critique from the 'Leaderboard Illusion' paper, defending LMArena's transparency and commitment to real-world usage data. He clarifies misconceptions about pre-release testing and asserts the platform's integrity in reflecting model performance based on user interactions.

The transition from Gradio to React is another key development for LMArena, driven by the need for enhanced performance and developer familiarity with React. This move is part of their broader strategy to provide a more robust and scalable platform.

Angelopoulos discusses the potential for LMArena to explore new categories and the possibility of launching video functionalities. They are focused on maintaining the integrity of their public leaderboard, which serves as a reliable benchmark for AI model performance.

Community engagement remains a priority, with strategies in place to retain and grow their user base. Angelopoulos emphasizes the importance of providing value to users, who are crucial for the platform's ongoing success and expansion.

Looking ahead, LMArena is open to partnerships and is actively seeking talented individuals to join their team. Angelopoulos invites collaboration with organizations interested in evaluating AI models and contributing to the platform's development.

Key Insights