AI benchmark evidence

Browse a selected MTEB embedding evaluation snapshot with source models, tasks, splits, languages and revision evidence kept separate.

MTEB evaluates embeddings, not chat completion quality. This is a selected, pinned snapshot, not the entire leaderboard. Compare the same task, metric, split, language subset and model revision. Missing declared primary scores are not treated as zero. Source: embeddings-benchmark/results, CC0-1.0.

Enable JavaScript for the interactive research workspace.