What is the Elo rating in Arena?

The Elo rating is a relative skill level calculation (originally from chess) that ranks models based on their win/loss ratio against other models in blind head-to-head battles.

Are the models in the Arena really anonymous?

Yes. During the battle, model names are hidden. Identity is only revealed after you submit your vote to ensure the evaluation is based purely on response quality.

Can I test specific models like GPT-4 or Claude 3.5?

Yes, using the 'Direct Chat' or 'Side-by-Side' mode, you can select specific models to test rather than relying on the randomized battle assignment.

Does it cost anything to use Arena?

No, Arena is a non-commercial, open-source research project and is completely free for the community to use and contribute votes.

Is there an official API for the leaderboard data?

There is no official supported API for production use, though historical datasets are released periodically and research tools are available via the FastChat GitHub repository.

Does Arena evaluate images and video?

Yes, the platform has expanded to include Vision Arena for image-based tasks and Video Arena for comparing generative video models.

Nobody owns this listing yet

Arena (formerly LMArena) is live in the directory, but no verified owner controls the page.

Claim this listing

Free to claim · 2 minutes

Pick your launch day — Schedule a launch and compete for the daily #1 spot.
Win trophies you can embed — Rank top 3 on launch day and get an embeddable trophy.
“Featured on AI Kaptan” badge — Owner-only badge with your live upvote count.
Control what the page says — Edit copy, pricing, links and screenshots as the owner.

Unclaimed listings don't stay up forever. Listings left unclaimed are removed from the directory in periodic cleanups. When a listing goes, it goes everywhere — the tool page and every alternatives comparison it appears on, along with the rankings and backlinks that page had built up.

Verify with a work email on the domain (about 2 minutes), then post about it on LinkedIn and tag us. See everything you get

AI model comparison

Free

Arena (formerly LMArena)

The gold standard for AI model rankings. Compare top LLMs through blind A/B testing and view the community-driven Elo leaderboard for text, vision, and coding models.

Visit WebsiteView Alternatives

—

Rating

Reviews

Upvotes

5 months ago

Listed

LLM benchmarking

Elo rankings

Model comparison

LMSYS

AI leaderboard

Blind testing

Open source evaluation

Multimodal AI

Vision models

Chatbot arena

Tool information

Provider

LMSYS Org

Platforms

Web

Languages

English

Chinese

Spanish

French

German

Japanese

Korean

Bengali

Catalan

Ukrainian

Added to directory

5 months ago

Last updated

2 months ago

Rating

Not available

Pricing

Free

About Arena (formerly LMArena)

What the tool does and who it's for

Arena (formerly LMArena) is the premier crowdsourced benchmarking platform for Large Language Models (LLMs), operated by LMSYS Org and Arena Intelligence. It utilizes a blind, pairwise evaluation system where users interact with two anonymous AI models simultaneously and vote on the superior response. These millions of human votes are processed using the Elo rating system to generate a real-time, authoritative leaderboard that ranks top-tier models from providers like OpenAI, Anthropic, and Google. The platform has expanded beyond text to include vision, coding, and video modalities, providing the AI industry with a transparent, qualitative signal of model performance.

Key capabilities

Blind A/B Testing: Compare two anonymous models side-by-side to eliminate brand bias in evaluations.

Elo Rating Leaderboard: Real-time rankings based on millions of crowdsourced human preference votes.

Multimodal Evaluation: Dedicated arenas for testing Vision-Language Models (VLM) and image-to-text capabilities.

Category-Specific Rankings: Filter leaderboards by Coding, Hard Prompts, Long Context, and specific languages.

Direct Chat Mode: Test specific known models (e.g., Llama 4, GPT-5) outside of the randomized battle format.

Vision Arena: Evaluate AI "eyes" by uploading images for OCR accuracy and spatial reasoning tasks.

Open Dataset: Access to anonymized prompt-response-vote triples for academic and safety research.

Video Arena: (New for 2026) Comparison platform for high-fidelity video generation and analysis models.

Adaptive Sampling: High-performing or new models are prioritized in battles to ensure rapid ranking convergence.

Transparent Methodology: Open-source infrastructure and public evaluation data to advance AI research.

Pricing

Free — plans below

Public Access

Reviews

Be the first to review Arena (formerly LMArena)

No reviews yet

Be the first to share your experience with Arena (formerly LMArena).

Featured Tools

Handpicked by our team of experts

Suggest an Update

Found outdated information? Help us keep this listing accurate.

FAQ

Arena (formerly LMArena) FAQs

Common questions about Arena (formerly LMArena)