Learn how your votes power transparent AI progress
Your conversations may be shared to support our community, improve our service, and advance the development of reliable AI. This includes posting conversations publicly online. Any data that we share is always anonymous and never linked to you. We never share any personal information, just the conversation and votes.
Your votes directly shape the model rankings through the Elo rating system, a method originally developed for ranking players in competitive games like chess. We use Elo because it's well-suited for pairwise comparisons, allowing us to update model scores incrementally based on real user preferences. The more you vote, the more reliable and representative the leaderboard becomes. This makes LMArena leaderboards grounded in community judgment, not static benchmarks.
We take user privacy seriously. All prompts and votes are anonymous and not connected to personally identifiable information. Additionally, individual conversations are never publicly shared beyond prompt text and model responses, ensuring your identity remains protected.
We openly collect prompts to maintain transparency, reproducibility, and trustworthiness of the evaluations. Open data helps the community independently verify model performance and ensures our benchmarks reflect authentic, real-world scenarios.
When you vote, you're directly influencing the public leaderboard rankings. Your choice updates each model's score using the Elo rating system, a method originally developed for ranking players in competitive games like chess. We use Elo because it's well-suited for pairwise comparisons, allowing us to update model scores incrementally based on real user preferences.
Yes, the models remain anonymous during voting to ensure fairness and eliminate potential bias. The model names are revealed immediately after you cast your vote, so you can discover which model you preferred. Note: Only votes made while the models are anonymous count toward official rankings; any votes cast after model identities are revealed will not impact leaderboard standings.
Absolutely! You're welcome to submit as many prompts and votes as you'd like. Every vote helps improve the accuracy and diversity of the leaderboard, making it more reflective of real-world model performance.
A few things to keep in mind:
We work directly with open-source and commercial model providers to make their pre-release models available for community testing, often before they appear anywhere else. This gives you early access to frontier models still in development, allowing you to explore, compare, and provide feedback while they're still being shaped. You may see these models appear under codenames or aliases in Battle mode. Before releasing a new model or version, AI companies test many variations to find the best one within their own closed doors. At LMArena, we make it open to everyone, so real-world feedback, transparency, and your voice can directly influence which models move forward.
Every model provider tests a differing variety of variants, depending on their needs. If a model passes both the company's criteria for public release and the requirements outlined in our policy, that particular model's score will be added to the public leaderboard under its official name.
Deep dive into our methodology by reading our research paper that explains the tried-and-true statistical methods we are using for efficient and accurate evaluation and ranking of models.
When you vote in Side by Side mode, you'll notice that you're selecting specific models, and the matchups are not anonymous. Votes in this mode do not contribute to the public leaderboard rankings. That said, your choices and prompts are still collected for research purposes. This helps the community, and the broader AI ecosystem, better understand human preferences. Prompts and votes from Side by Side mode support our mission to make AI progress more transparent by sharing open research. See some of that research on our blog here.
When you interact with specific models in Direct Mode, you'll notice there is no voting. Similarly to Side by Side, your prompts are collected for research purposes. This helps the community, and the broader AI ecosystem, better understand human preferences. See some of that research on our blog here.
Yes! We share a portion of our anonymized voting data with the research community to support open science and reproducibility. While we don't release full conversation logs for privacy and methodological reasons, the available data includes prompt text, voting outcomes, and model pairings. You can explore our datasets with Arena Explorer or directly on HuggingFace and reach out if you're interested in collaborations or deeper access.
LMArena is powered by a diverse global community, from AI enthusiasts and students to researchers, developers, and everyday users. Everyone is welcome to this open space, because we believe the best AI evaluations reflect real-world diversity and lived experience, not just expert opinion.
As of May 2025, LMArena is backed by Andreessen Horowitz (a16z) and UC Investments (University of California), with additional participation from Lightspeed, Laude Ventures, Felicis, Kleiner Perkins, The House Fund, and others.
Learn more in our LMArena and The Future of AI Reliability blog post.