A focused, multilingual embedding specialist that converts text into dense vector representations across many languages. It operates within a tight 514-token window, so it works best with shorter passages rather than long documents. Its open-weight nature means it can run locally without API dependencies.