Harrier tops the multilingual MTEB v2 benchmark. The model, open‑sourced by the Microsoft Bing team, supports more than 100 languages. Its broad coverage and high performance make it a ready‑to‑use foundation for multilingual NLP tasks. Developers can integrate Harrier into pipelines without retraining, accelerating deployment. Because it is open‑source, teams can fine‑tune it on domain data if needed.