LMArena Leaderboard
Crowd-sourced head-to-head preference Elo rating. · Prices checked November 11, 2026
LMArena rates models from blind head-to-head comparisons: a user sends one prompt, sees two anonymous responses, and picks the better one. Those votes feed an Elo rating of the kind used in chess.
Leaderboard
60 of 149 scoredLeaderboardScores as published by each benchmark. Prices checked November 11, 2026.
Claude Opus 4.6 leads on LMArena Elo at 1497 Elo. The median across the 149 models we hold a score for is 1303 Elo. The cheapest model on this board is Yi-Lightning at $0.1400/M blended, scoring 1328 Elo.
60 of the 60 scores shown link to the page they were read on (underlined; hover for publisher, variant and date read). The rest have no recorded source yet and are being worked through.
Because it aggregates human preference over open-ended prompts rather than performance on a fixed test set, it captures general helpfulness in a way scored benchmarks cannot, and it is correspondingly harder to game.
About these scores
- Crowd-sourced head-to-head preference Elo rating.
- Scores are as published by the model’s provider or a public leaderboard. Provider-reported figures are self-reported and are not independently re-run by us.
- Where a hosted variant of an open-weights model has no score of its own, it inherits the base model’s, because the weights are the same.
- Prices beside each score are the provider’s published token rates. How the blended rate is calculated.
Showing the top 60 of 149 models we hold a LMArena Elo score for.
Frequently asked questions
What is LMArena Elo?
A rating derived from blind pairwise votes. Users compare two anonymous model responses to the same prompt and choose the better one; the winner gains rating points and the loser gives them up, exactly as in chess. A higher Elo means users preferred that model more often across many head-to-head matchups.
Why does LMArena disagree with other benchmarks?
It measures a different thing. Scored benchmarks test correctness on problems with known answers; LMArena measures which response people prefer on open-ended prompts, which rewards tone, formatting, instruction-following and usefulness. A model can be excellent at graduate physics and still lose votes for being terse or badly formatted.
Other leaderboards
Rankings that weigh price too
Leaderboards are recomputed on every deploy from the catalogue. Prices checked November 11, 2026.