Cohere embedding models

Cohere delivers industry leading text embeddings with specialized support for 100 plus languages and native RAG optimization.

Cohere's Embed v3 models transform unstructured text into high dimensional vectors for semantic search and retrieval augmented generation (RAG). The architecture features a unique compression technique that reduces storage costs by up to 100x while maintaining 99% of original search accuracy. Developers access these models via a single API endpoint to power applications like enterprise search, content recommendation, and anomaly detection. With support for over 100 languages (including English, Spanish, and Mandarin), Cohere consistently ranks at the top of the MTEB leaderboard for retrieval performance and document ranking.

https://cohere.com/embeddings