Arena

Interesting sitesLeaderboardLevel: EasyFree plan
Logo Arena

Arena

4.0/52dm rating

A model leaderboard based on real votes; worth checking before choosing a model.

Arena (formerly LMArena, Chatbot Arena) ranks AI models based on user votes from anonymous side-by-side comparisons.

Overview

Arena is a platform that evaluates AI models using votes from real users. The project started as Chatbot Arena at UC Berkeley, was later renamed LMArena, and from 28 January 2026 officially took the name Arena, moving to arena.ai.

How it works is simple: in Battle mode, you enter a request, the system shows answers from two anonymous models, you pick the better answer, and only then are the model names revealed. These votes feed public leaderboards for many categories such as Text, WebDev, Vision, Document, Search and Agent, as well as image and video generation and editing categories.

For businesses, Arena helps answer the question "which model do users currently rate highly for this kind of task" and lets you compare models yourself with your own questions before choosing a paid tool.

Highlights

01

Anonymous comparison

Two models answer the same request without you knowing their names, making the judgment more objective.

02

Leaderboards in many categories

Overall, Text, WebDev, Vision, Document, Search, Agent and image and video categories.

03

Side-by-side comparison with model selection

Besides Battle, you can pick models yourself and compare them side by side.

04

Based on real votes

Rankings reflect the choices of many users on real tasks.

05

Update log

A Leaderboard Changelog lets you track when the rankings change.

What it's for

Choose a chatbot for the team

Before buying AI plans for staff, a manager checks the Text category to see which model is leading.

Choose a model for coding or web work

Technical teams consult the WebDev category when choosing a coding assistant.

Choose image or video ad tools

Marketing looks at the Text to Image, Image Edit and Text to Video categories to shorten the tool list.

Test with your own questions

Staff use Battle with a sample question (with internal information removed) to see which model answers better in Vietnamese.

Quick updates when new models appear

The person in charge of AI follows the leaderboard to see whether a new model is really better.

How to use it

  1. 1

    Open the leaderboard

    Go to arena.ai/leaderboard and choose Overall for an overview.

  2. 2

    Choose the right category

    Switch to Text, WebDev, Vision, Text to Image or Text to Video depending on the kind of work you need.

  3. 3

    Try Battle mode

    Click Start Voting, enter a request, read the two anonymous answers and pick the better one to see the model names.

  4. 4

    Compare models you care about

    Use the model selection or side-by-side feature to put two specific models next to each other.

  5. 5

    Cross-check before deciding

    Combine the ranking with the provider's price, speed and data terms before choosing.

Video guides

Behind LMArena's leaderboard: understanding AI model performance

Arena AI

LMArena statistical ranking method explained - Ask an Engineer

Arena AI

Discover the Best AI Models Using LM Arena Platform EP #214

Everyday AI

Real-world example

Illustrative scenario

A small real estate company wants to choose an AI assistant to write apartment descriptions in Vietnamese. Instead of choosing by advertising, a sales employee goes to Arena, looks at the Text category, then uses Battle mode with a few sample requests that contain no customer information. After seeing which group of models writes more natural Vietnamese, the company compares the prices of the corresponding paid plans.

Tips & notes

  • Don't enter sensitive data

    According to the How it works page, prompts are collected for public evaluation and model development.

  • Choose the right category

    The model leading Text does not necessarily lead WebDev or image generation.

  • Verify for Vietnamese yourself

    The overall ranking does not specifically reflect Vietnamese quality, so try your own questions.

  • Combine multiple sources

    Rankings based on user preference should be cross-checked against other benchmarks such as Artificial Analysis.

FAQ

Are LMArena and Arena the same?

Yes. Chatbot Arena was renamed LMArena, then on 28 January 2026 renamed Arena at arena.ai.

How are the rankings produced?

From user votes when comparing answers of two anonymous models; see the official site for method details.

Does using Arena cost money?

The leaderboard is publicly viewable; see the official site for the terms of other features.

Should I enter company documents to test?

No, because prompts are collected for evaluation purposes; use only sample content.

Useful links

Want to apply this in your business?

2dm looks at your workflows, picks the right tools and sets them up with your team. The first call is free.

Book a call
Zalo