In the early hours, Google released its first embedding model, gemini-embedding-001, which immediately topped the MTEB leaderboard, surpassing OpenAI

Google has released its first embedding model, gemini-embedding-001, in the early hours, quickly claiming the top spot on the MTEB benchmark leaderboard—surpassing OpenAI. With an average score of 68.37 on MTEB, it significantly outperforms OpenAI’s text embedding model, which scored 58.93. Priced at just $0.15 per million tokens, the model excels in tasks such as bilingual mining, classification, clustering, instruction retrieval, multi-label classification, pair classification, re-ranking, retrieval, and semantic textual similarity. It supports 100+ languages and handles input lengths of up to 2048 tokens.