Search
On June 6, Alibaba unveiled a new series of vector models, Qwen3-Embedding, tailored for Tongyi Qwen3. Built upon the robust foundation of the Qwen3 base model, these models are specifically engineered for text representation, retrieval, and ranking tasks. Notably, they deliver a remarkable 40% performance enhancement compared to their predecessor, outperforming top models from Google, OpenAI, Microsoft, and other industry leaders in the MTEB multilingual benchmarks. This achievement places Qwen3-Embedding at the forefront of state-of-the-art (SOTA) performance among similar models. The Qwen3-Embedding series comes in three configurations, ranging from 0.6B to 8B parameters, supporting over 100 languages and boasting robust multilingual processing capabilities. This offers developers efficient and user-friendly solutions tailored to their needs.
-
C114 Communication Network -
Communication Home
