Google has introduced Gemini Embedding, a new experimental embedding model that generates numerical representations of text for improved applications like document retrieval and classification. This marks the first embedding model from Google utilizing its Gemini AI models. Google highlights that Gemini Embedding not only offers better performance than the previous text-embedding-004 but also accommodates larger text inputs and supports over 100 languages. Currently in an experimental phase, the model has limited capacity but is expected to stabilize for a broader release in the coming months.
Embedding models translate text inputs like words and phrases into numerical representations, known as embeddings, that capture the semantic meaning of the text.
Trained on the Gemini model itself, this embedding model has inherited Gemini's understanding of language and nuanced context making it applicable for a wide range of uses.
Gemini Embedding surpasses the performance of its previous state-of-the-art embedding model, text-embedding-004, and achieves competitive performance on popular embedding benchmarks.
Gemini Embedding is in an 'experimental phase' with limited capacity and is subject to change, with a stable release expected in the months to come.
Collection
[
|
...
]