Qwen3 Embedding is a series of embedding and reranking models designed for text retrieval, classification, clustering, and bitext mining. The models inherit multilingual capabilities and long-text understanding from the Qwen3 foundation models. Available in 0.6B, 4B, and 8B parameter sizes, they achieve state-of-the-art performance on MTEB benchmarks, ranking #1 on the multilingual leaderboard as of June 2025.