State-of-the-art text embedding via the Gemini API- Google Developers Blog

What this post added

Introduced a new experimental Gemini Embedding text model (gemini-embedding-exp-03-07) available via the Gemini API. This model surpasses previous state-of-the-art text embedding models, achieving top rank on the MTEB Multilingual leaderboard. Key features include an input token limit of 8K, output dimensions of 3K, Matryoshka Representation Learning (MRL) for dimension scaling, and expanded language support to over 100 languages. The model is trained on the Gemini model itself, inheriting its language understanding and contextual capabilities.