Arena (formerly Chatbot Arena)
Arena (formerly Chatbot Arena) is a web-based platform for comparing AI model responses and building public leaderboards from votes.
What Arena (formerly Chatbot Arena) is
Arena (formerly Chatbot Arena) is a web-based platform for comparing AI model responses and building public leaderboards from votes. In Battle mode, users see two anonymous responses and learn the models’ identities after voting. Arena aggregates eligible votes with the Bradley-Terry rating system; only anonymous Battle votes count toward official rankings. Side by Side and Direct modes are also available. Rankings cover agent, text, web development, image, video, vision, document, and search tasks, and comparisons include image generation, image editing, and video generation. Leaderboards can be viewed without signing in, while Direct, Side by Side, Image Arena, and Video Arena require login. Arena says conversations may be used in research datasets and private model-provider evaluations; shared data is de-identified, and users are advised not to submit sensitive information. A daily usage limit applies to everyone and resets 24 hours after it is reached. Video Arena requires login, allows two generations per 24-hour period, and supports Battle mode only. The free Public Arena plan is free. Commercial evaluation services for enterprises, model labs, and developers are also offered.
Good
- Free public leaderboards and head-to-head battles
- Leaderboards can be viewed without signing in
- Rankings span text, image, video, and other tasks
- Votes are aggregated using the Bradley-Terry system
Watch out
- Daily usage limit resets 24 hours after it is reached
- Video Arena allows two generations per 24-hour period
- Several modes require login
- Conversations may be used for research or evaluations
Who it suits
Arena suits people who want to compare model responses or browse public rankings across different task types. It also offers commercial evaluation services for enterprises, model labs, and developers.
Verdict
Arena offers a free way to compare model responses and explore public rankings, with limits and login requirements for some features. Avoid submitting sensitive information, since conversations may be used in datasets or provider evaluations.
Plans and prices
From Arena (formerly Chatbot Arena)'s own pricing page, checked 7 Oct 2026.
| Plan | Price | What it gives you |
|---|---|---|
| Public Arena | Free | Public leaderboard and head-to-head battles are free; daily usage limit applies and resets after 24 hours |
| AI evaluations | Priced on request | for enterprises, model labs and developers · grounded in real-world human feedback |
The facts
Each one read from Arena (formerly Chatbot Arena)'s own pages, with the date it was checked.
What it does
- Account benefits
- Creating an account unlocks Direct, Side-by-Side, Image, and Video modes, higher rate limits, and synchronized chat history.checked 21 Oct 2026
- Account integrations
- Users can create an account with Google or email, and Arena’s coding Agent Mode can connect to GitHub repositories.checked 7 Oct 2026
- Account requirements
- Leaderboards can be viewed without signing in, while Direct, Side by Side, Image Arena, and Video Arena require login.checked 7 Oct 2026
- Agent files
- Agent Mode accepts PNG, WebP, JPEG, PDF, GIF, TXT, Markdown, CSV, HTML, XML, CSS, JavaScript, and JSON uploads.checked 21 Oct 2026
- Agent mode
- Agent Mode plans multi-step work and can use web search, image generation, file uploads, coding assistance, and a sandbox/bash environment.checked 21 Oct 2026
- Age restriction
- The privacy policy says the service is not intended for people under 18.checked 21 Oct 2026
- Anonymous battles
- In Battle mode, Arena presents two anonymous model responses and reveals their identities after the user votes.checked 7 Oct 2026
- API
- Arena provides an OpenAI-compatible gateway with model listing and chat-completion endpoints authenticated by virtual API keys.checked 21 Oct 2026
- API compatibility
- The API documentation shows Anthropic-compatible client configuration for Claude Code through the Arena API base URL.checked 21 Oct 2026
- Battle mode
- Users submit a prompt, receive two anonymous model answers, vote, and then see the model identities.checked 21 Oct 2026
- Commercial customers
- Arena offers commercial evaluation services to enterprises, model labs, and developers, with evaluation results supported by service-level agreements.checked 7 Oct 2026
- Data sharing
- Arena says conversations may be de-identified and shared in public research datasets or private evaluations with model providers.checked 21 Oct 2026
- Developer documentation
- The public API reference identifies itself as OpenAPI 3.0.3 and includes curl quickstarts.checked 21 Oct 2026
- Discord
- The terms cover an Arena Discord server and an Arena Discord Bot as part of the service.checked 21 Oct 2026
- Funding
- Arena says it has $250 million in venture funding led by Felicis and UC Investments, with participation from several other investors.checked 21 Oct 2026
- GitHub integration
- When a GitHub repository is connected, Agent Mode works in a copy, commits changes to a branch, and opens a pull request.checked 21 Oct 2026
- Independent research result
- The Chatbot Arena paper reports more than 240,000 votes and says crowdsourced questions were diverse and human votes agreed well with expert ratings.checked 21 Oct 2026
- Licence status
- The LMSYS project describes the FastChat platform, including its UI, serving backend, evaluation, and ranking pipelines, as open source and runnable by others.checked 21 Oct 2026
- Limit: benchmark gaming
- Contrary Research reports that labs deliberately target Arena and that self-selected, unpaid voters and test-set exposure can make results gameable.checked 21 Oct 2026
- Limit: directory-only review
- Aigregator says the similarly named listing is mainly a discovery and reference tool, requires visiting builders for hands-on testing, and may contain outdated comparison data.checked 21 Oct 2026
- Limit: model mismatch
- Contrary Research describes the 2025 Llama 4 Maverick episode in which a tuned Arena variant ranked highly while the publicly released model ranked much lower.checked 21 Oct 2026
- Limit: outages
- A Russian outage monitor records historical LMArena complaints involving site access, app access, photo loading, and home-network availability.checked 21 Oct 2026
The company
- Company origin
- Arena says it was created by UC Berkeley researchers and that Arena Intelligence Inc. was formed in 2025.checked 7 Oct 2026
- Founded
- 2023checked 28 Sep 2026
Alternatives to Arena (formerly Chatbot Arena)
- 1
DeepEval
Free planOpen sourceWebAPILinuxmacOS8.3From $200 a month - 2
Promptfoo
Free planOpen sourceWebAPILinuxmacOS7.8Free plan - 3
Braintrust
Free planWebAPISelf-hosted7.7From $249 a month - 4
Giskard
Free planOpen sourceWebAPILinuxSelf-hosted7.6Free plan - 5
TruLens
Free planOpen sourceAPISelf-hosted7.2Free plan - 6
Maxim AI
Free planWebAPISelf-hosted7.0From $29 a month