INSTRUCTOR

A state-of-the-art multilingual text embedding model from Alibaba's GTE (General Text Embedding) series, built on the Qwen2-1.5B LLM. The model supports up to 8192 tokens and incorporates bidirectional attention mechanisms for enhanced contextual understanding across diverse domains.

Open-source embeddings model from Microsoft initialized from Mistral-7B-v0.1, achieving state-of-the-art BEIR score of 56.9 for English text embedding and retrieval tasks with 4096-dimensional vectors.

A versatile multilingual text embedding model from BAAI that supports 100+ languages and can handle inputs up to 8192 tokens. BGE-M3 is unique in supporting three retrieval methods simultaneously: dense retrieval, multi-vector retrieval, and sparse retrieval.

A large-scale multilingual text embedding model from Alibaba's GTE series with 7 billion parameters. Built on Qwen2-7B, it achieved a score of 70.24 on MTEB, outperforming NV-Embed-v1 and supporting 100+ languages with up to 8192 token context.

A family of English text embedding models distilled from state-of-the-art embedding models using a novel multi-stage distillation framework. Stella models support multiple dimensions (512 to 8192) through Matryoshka Representation Learning, offering flexible embedding sizes for different use cases.