Home › Benchmarks › LMArena (Chatbot Arena)
LMArena (Chatbot Arena)
Benchmark
Capability
Millions of blind human votes — preference as a leaderboard.
You're looking at a benchmark record: a test the industry uses to measure chips or models — and what it actually tells you.
LMArena shows users two anonymous answers and asks which is better, aggregating millions of votes into Elo ratings. It measures what benchmarks cannot — perceived quality — and its weaknesses are equally human: style bias, and vendors tuning for votes (the Llama 4 episode made this a governance story).
How this connects to the rest of the stack
This record's links at a glance — click a group to jump to it.
Benchmarks
Sources
Record reviewed as of 2026-01-15 — see Methodology.