LLM Leaderboard

How today's language models compare, by LMArena's overall rating: a score from crowdsourced head-to-head votes, where people compare two anonymous answers and pick the better one. The list is LMArena's current top 50, taken from its published data and checked for updates every 6 hours.

LMArena doesn't record architectures. Closed models are marked "Undisclosed"; an architecture appears only where it is recorded with a source. For how these models are built, see the catalog.