Model4d ago
Bing Open‑Sources Harrier, 100‑Language Embedding
Harrier tops the multilingual MTEB v2 benchmark, scoring 1,200 points. The model, released by Microsoft’s Bing team, supports more than 100 languages, covering a broad swath of global text. By open‑source, Bing gives developers ready‑to‑use embeddings for search and translation. Practitioners can now integrate multilingual embeddings without training from scratch.