Arena AI

The world's largest AI model crowdsourcing evaluation platform, user voting ranking

List Tracking

Model Evaluation

Model Evaluation Platform

Arena AI, formerly known as Chatbot Arena and LMArena, is a crowdsourced AI model evaluation platform created in 2023 by a research team at the University of California, Berkeley. The platform adopts a unique blind test mode. After users input prompt words, two anonymous AI models respond at the same time. After users vote for better answers, the system reveals the identity of the model. Through millions of real user votes, an authoritative ranking list based on ELO scoring system is generated. The platform has become the most concerned evaluation standard in the AI industry, and top labs such as OpenAI Google Anthropic submit unreleased models to Arena for testing. In January 2026, the company completed Series A financing of US$150 million, with a valuation of US$1.7 billion. Investors included Felicis, University of California, Andreessen Horowitz, and other well-known institutions. With more than one million monthly visitors, Arena Intelligence Inc has transformed from an academic project to an independent company, headed by postdoctoral researcher Anastasios Angelopoulos as CEO, promising to remain neutral and open and continue to provide transparent AI evaluation services to the community.

Preview

Introduction

functional characteristics

Arena AI It does not provide text, voice, image or video generation capabilities, but rather a platform focused on evaluating and comparing the performance of other AI models.
Its core function is to allow users to vote on the response of two AI models to the same prompt word without knowing the identity of the model through a unique blind test mode, so as to objectively evaluate the model.

application scenarios

AI model development: Top labs (e.g. OpenAI, Google, Anthropic) submit unpublished models here for stress testing and performance validation.
Technology Selection Reference: Enterprises and developers use their authoritative leaderboards to select the most appropriate AI model for their projects.
Academic research: Researchers use their ELO score data, based on millions of real user votes, to conduct AI model performance analysis and trend studies.
Industry standard setting: As the most concerned evaluation benchmark in the AI industry, it provides transparent and credible model capability measurement for the whole ecosystem.

suitable for the croud

AI researchers, engineers and product managers.
Enterprise decision makers and technical teams who need to choose the right AI model for their business.

type of pricing

The platform is free and open to users, and its business model relies primarily on venture capital support rather than user fees.

supported langauges

The platform interface and evaluation process are primarily English-based, but the evaluation model itself may support multiple languages.

technical highlights

Innovative Blind Test Match (LLM) Arena) mechanism and reinforcement learning based on human feedback (RLHF) concept, quantifying model performance through ELO scoring system.
The results have become a de facto standard for measuring the capabilities of large models, with impact comparable to official benchmarks and even more reflective of the real user experience.

usage mode

The Arena AI platform (lmsys.org) can be accessed directly through a web browser for model battle experiences and leaderboards.

update history

The platform continues to iterate its evaluation methodology and leaderboard algorithms to adapt to rapidly evolving AI model capabilities, and formally transforms from an academic project to an independent company Arena after completing Series A funding in early 2026. Intelligence Inc.。
Recent updates focus on scaling up reviews, optimizing the user experience, and promising to remain neutral and open, continuing to provide transparent AI review services to the community.