LMArena Secures $150M to Redefine AI Model Evaluation
LMArena: Redefining How the World Evaluates Artificial Intelligence
The artificial intelligence industry has grown at a pace that has far outstripped its ability to measure itself. Hundreds of large language models compete for attention, yet the traditional benchmarks used to rank them are often static, contaminated by training data, or produced by the very companies whose models are being evaluated. Into this gap stepped LMArena — a platform built on a simple but powerful idea: let real people decide which AI is best.
From Academic Project to Billion-Dollar Company
LMArena was originally launched by the UC Berkeley Large Model Systems Organization (LMSYS) as an academic side project. What began as a research experiment quickly became one of the most trusted tools in the AI industry. In April 2025, LMArena incorporated as an independent company, and that May, it raised $100 million in a seed funding round, valuing the company at $600 million. By the start of 2026, its growth had accelerated dramatically, with a new $150 million funding round pushing its valuation to $1.7 billion — nearly triple what it had been just eight months earlier.
The Founding Team
LMArena was formally founded by Anastasios N. Angelopoulos as CEO, Wei-Lin Chiang as CTO, and Ion Stoica as Co-founder and Advisor, with the mission to build a more rigorous and transparent foundation for evaluating large language models. Angelopoulos brought expertise in trustworthy AI systems and medical machine learning, while Chiang contributed deep experience in distributed systems and deep learning frameworks. Together, they assembled a team with roots at Google, DeepMind, Discord, Vercel, Berkeley, and Stanford.
How the Platform Works
LMArena operates as a neutral benchmarking platform that enables users to compare large language models through head-to-head matchups. It works by allowing users to submit prompts and evaluate anonymous responses from different models, selecting the best reply. The identities of the two competing models are revealed only after the user votes. This blind, crowdsourced approach produces rankings that reflect genuine real-world preferences rather than engineered scores. Unlike traditional static benchmarks, which are often vulnerable to data contamination from model training, LMArena employs a direct comparative system where users rate responses in real time, providing a dynamic performance metric that reflects practical utility in tasks such as text generation, web development, and complex reasoning.
Scale and Community
The platform’s growth in user participation has been remarkable. LMArena’s community now spans more than 5 million monthly users across 150 countries, who collectively generate more than 60 million conversations every month. Over four hundred model evaluations have already been completed on the platform, with over 3 million votes cast, helping shape both proprietary and open-source models across the industry, including those from Google, OpenAI, Meta, and xAI.
Industry Adoption and Commercial Growth
LMArena has become indispensable to the world’s leading AI laboratories. It works closely with leading AI labs and enterprises, including OpenAI, Google, and xAI, all drawing on LMArena’s evaluations to improve their models for production use cases and user preferences. The platform has also expanded into commercial territory, earning revenue by providing paid AI evaluation services across economically valuable industries like software engineering, law, medicine, and scientific research. Its first commercial product launched in September 2025, and the company’s annualized revenue run rate surpassed $30 million in December 2025, less than four months after launch.
The Vision Ahead
What started as a Berkeley research project has quickly become essential infrastructure — a continuous integration pipeline for intelligence — not because of marketing or sales, but because the platform solved a problem everyone had but no one addressed. As AI increasingly enters high-stakes domains like healthcare, law, and government, the need for neutral, reproducible evaluation has never been greater. With a commitment to open access, reproducible methods, and diverse human judgment, LMArena is building the foundation for the world to understand, shape, and benefit from AI. In an industry where trust is everything, LMArena has become the standard-bearer for accountability — and its story is only just beginning.
