Google tests Gemini embedded model

On March 7, Google announced that it has added a new experimental embedding model for text, called Gemini Embedding, through the Gemini Application Programming Interface API.

Google announced that it has added a new experimental embedding model for text, called Gemini Embedding. (Source: Getty Images)

In a blog post, Google emphasized that the experimental embedding model, trained on the Gemini model itself, inherits Gemini’s ability to understand language and nuanced context. The new model also has “incredible” generalization capabilities, delivering superior performance across a wide range of domains including finance, science , law, search, and more.

Google notes that Gemini Embedding is still in beta, so there are still capacity limitations. The company aims to continue to refine and improve it so that it can release a stable version to a wider audience in the coming months.

In recent times, companies including Amazon, Cohere, and OpenAI have provided embedding models through their respective APIs. Gemini Embedding is Google's first embedding model trained on the Gemini AI model.

Embedding models are algorithms trained to encapsulate information into dense representations in multidimensional space.

Data scientists use embedding models to help machine learning models understand and reason with multidimensional data.

Source: https://baoquocte.vn/google-thu-nghiem-mo-hinh-nhung-gemini-306828.html