A specialist that does one thing and does it well: turning text into dense numerical representations for search and retrieval tasks. It's a workhorse embedding model — not conversational, not generative — just focused on producing high-quality vector embeddings. The 512-token context window means it works best with shorter passages rather than long documents.