Model4d ago
Microsoft Open‑Sources Harrier Embedding Model
Harrier tops the multilingual MTEB v2 benchmark, scoring 1,200 points across 100+ languages. Developed by Microsoft's Bing team, the open‑source embedding model offers developers a ready‑to‑use multilingual foundation that simplifies cross‑lingual tasks from search to translation. Teams can integrate Harrier into pipelines without training from scratch, and the release encourages community contributions that may accelerate multilingual AI research.