qwen3-embedding:4b

Family context: Qwen3 Embedding family

Publisher: Qwen Team

Qwen’s embedding releases convert text into vectors for retrieval, similarity and classification workflows. They are separate from the family’s reranking checkpoints and from chat-generating models.

Publisher-described context, separate from Sigma5C observations and recorded catalog statistics.

Published family information

Published embedding sizes: 0.6B, 4B and 8B Source
Published sequence length: 32K tokens Source
Vector dimensions by size: 0.6B:1024; 4B:2560; 8B:4096 Source
Dimension adaptation: Matryoshka-style truncation is documented Source

Publisher-described uses

Text and code retrieval, classification, clustering and cross-language text matching. Source

Reference limitations

The embedding and reranking products have different outputs; do not label an embedding model as a text-generation or reranking model. Source

A recorded tag and an upstream card association do not verify the deployed checkpoint, quantization, configured context or current hosted availability.