ULTRABENCH needs JavaScript for the interactive leaderboard

The data does not. Every number in this index is served as JSON from a public, key-free API — start at /v1/leaderboard for the ranked table, or /v1/snapshot for the whole published index, /v1/models, /v1/benchmarks, /v1/families and /v1/hardware for the individual collections.

AlpacaEval 2.0

AlpacaEval 2.0: every model scored on it, ranked, with a source on every score, and the current models not yet covered.