Arena Coding

LMArena subcategory ranking models on user pairwise votes restricted to coding prompts.

Operator: LMArena
Kind: Human preference
Updates: weekly·updated 8d ago
Notable for: The reference public ranking for code-LLM preference, complementing execution-based benchmarks like SWE-bench and LiveCodeBench.
URL: lmarena.ai/leaderboard
Tracks: 344 models · Elo

Cite

Notes

Only stored in your browser.

Attribution

Scores: lmarena.ai/leaderboard

Attribution policy →

Elo ranking

344

models

Arena Coding · EloBar chart with 21 bars. Highest value: Claude Opus 4.6 at 1535.

21 models

Backing benchmark

Human-preference voting. No underlying benchmark - models are ranked by pairwise votes, not by a test you can run.