Qwen3-Embedding

Qwen3-Embedding is a family of models for developers building search and text analysis on their own hardware or servers. It represents text as numerical vectors so applications can compare passages by meaning.

• Multilingual and code search: Retrieve documents and code, including matches between queries and passages written in different languages, with support for English, Chinese, Arabic and Ukrainian.
• Text analysis: Use embeddings to classify text, group related passages or identify corresponding text across languages.
• Custom embeddings: Choose output vector dimensions and supply task instructions to adapt embeddings to a retrieval problem or language.
• Long documents and reranking: Process long text and pair the embeddings with companion Qwen3-Reranker models to reorder search results by relevance.

The family builds on Qwen3 dense base models and includes Qwen3-Embedding-0.6B, Qwen3-Embedding-4B and Qwen3-Embedding-8B. Local inference works through Sentence Transformers, Transformers and vLLM. Hugging Face Text Embeddings Inference supports self-hosted Docker deployment on CPUs or NVIDIA GPUs.

The Qwen3-Embedding-8B weights use Safetensors and carry an Apache 2.0 license.