A quiet, multilingual workhorse focused entirely on turning text into meaningful vector representations across languages. It handles cross-lingual semantic similarity without fanfare — sentences in different languages that mean the same thing land close together in its embedding space. As a base-sized model, it trades raw capacity for speed and accessibility.