ViDense is a focused, single-purpose specialist — it takes Vietnamese text and turns it into dense vector embeddings, nothing more, nothing less. It works quietly in the background of search and retrieval pipelines, converting sentences into numerical representations that capture semantic meaning. Its compact 8K context window keeps it lightweight and practical for most document-level tasks.