A specialized AI model trained specifically to convert text into vector embeddings, distinct from the LLM used for generation.
RAG systems typically use a separate, smaller embedding model for retrieval and a larger LLM for generation — each optimized for its specific job.
