MTEB Leaderboard

Choosing the right text embedding model used to mean running your own benchmarks or trusting marketing claims. This leaderboard removes that guesswork by ranking hundreds of embedding models across a comprehensive suite of real-world tasks including retrieval, clustering, classification, reranking, and semantic textual similarity. Each model is scored across multiple languages and task types, so you can filter by what actually matters for your use case rather than chasing a single headline number. For anyone building search, recommendation engines, RAG pipelines, or anything that depends on how well text gets vectorized, this is the reference you want bookmarked. It surfaces both the obvious giants and smaller, faster models that punch above their weight for specific domains, which matters enormously when you are balancing latency and cost against quality. The honest reservation is that benchmark performance does not always translate directly to your specific data distribution, so treat top scores as a strong starting point rather than a final answer. -> Best for: founders and engineers building semantic search, RAG systems, or any NLP product where embedding model selection directly impacts product quality.