A Generative AI Engineer is comparing two embedding models for a semantic search system that powers a legal assistant. Model X supports 512-token context and has a 384-dimensional embedding size Model Y supports 2048-token context and has a 1024-dimensional embedding size The assistant must retrieve highly relevant clauses from large legal documents, but inference latency and cost are constraints. The engineer needs a balance between retrieval depth and performance. Which model should the engineer select, and why?