Qwen3 Embedding
Open timelineWhat changedReleased
Qwen3 Embedding and Reranker
Qwen3 releases embedding and reranking models for multilingual retrieval.
- 0.6B, 4B and 8B variants use Apache 2.0.
Access at announcement
These models return vectors or relevance scores, rather than chat answers.
Announcements & evaluations
2 sourcesOfficial announcementsOfficial
Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation ModelsModel cards & docsOfficial
Qwen3-Embedding-8B · Hugging Face