AlibabaQwen Embedding
Qwen3 VL Embedding
Qwen3 VL Embedding supports fused and independent vectors with up to 32,000 input tokens.
Practical use
Practical guidance. Decide whether the query should match a whole image-text pair or one component. Keep that choice consistent during indexing and search, then inspect failures involving small visual details.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API |
| Model identifiers | qwen3-vl-embedding |
| Dimensions | 256–2560; default 2560.Embedding and reranking model catalogue |
Related models
- Qwen3-Embedding-8B
Alibaba · Qwen Embedding
Eight-billion-parameter multilingual embedding model with task instructions and a 32,000-token context.
Launch date not confirmedDownloadable - Qwen3-Embedding-4B
Alibaba · Qwen Embedding
Four-billion-parameter text embedding model for instruction-aware retrieval across more than 100 languages.
Launch date not confirmedDownloadable - Qwen3-Embedding-0.6B
Alibaba · Qwen Embedding
A 600-million-parameter text embedding model with task instructions, over 100 languages and a 32,000-token context.
Launch date not confirmedDownloadable