What is the MTEB Leaderboard?

The MTEB (Massive Text Embedding Benchmark) Leaderboard, hosted by Hugging Face, is the authoritative ranking of text-embedding models. It continuously evaluates thousands of public and private models on 100+ datasets across 8 task families: retrieval, reranking, semantic similarity, classification, clustering, summarization, pair classification and STS.

Key Features

• Multi-metric scoring: standardized nDCG@10, MAP, Spearman, etc.
• One-click reproduction: every row links to a runnable notebook and YAML config.
• Side-by-side comparison: tick any models to generate comparative charts.
• Live updates: new results from the community are integrated within 24 h.

Target Users

• Researchers: discover SOTA models and reproducible baselines.
• Engineers: quickly shortlist production-grade models by task.
• Students: learn how architecture, size and training data affect performance.

Quick Start

1. Use the top filters to narrow down by task, language or model size.
2. Click a model name to open its Hugging Face Model Card for details and download.
3. Hit the “Submit” button to upload your own evaluation and become a contributor.