Qwen3-Embedding-0.6B

Qwen3-Embedding-0.6B is a compact 0.6-billion-parameter multilingual text embedding model from Alibaba's Qwen team, designed for semantic search, retrieval, and RAG pipelines across 100+ languages with a 32K context window. It supports instruction-aware embeddings and Matryoshka Representation Learning, enabling flexible output vector dimensions from 32 to 1024. Source: supabase.