Documentation

At mixedbread, we are pushing the state-of-the-art further for a new era of language, retrieval-augmented generation (RAG), and search systems. Our product range includes our flagship embedding model, best-in-class reranking, and ColBERT, as well as experimental architectures like 2D-Matryoshka. We are constantly working on baking new and improved models for our community, so stay tuned for more exciting updates!

What is neural search?

In contrast to traditional keyword search, neural or semantic search involves searching for data based on its actual semantic and contextual relevance to your query. For search systems to work with this level of understanding, they need to access data in a way that captures the relevant semantic connections.This is where embeddings come into play.

Embedding models, such as neural networks or transformers, convert complex data (text, images, video, tables, audio) into numerical vectors, capturing semantic relationships. This technology is crucial for developing sophisticated search systems and retrieval-augmented generation (RAG) stacks, elevating chatbot functionality, search capabilities, recommendation systems, and much more across various industries and enterprises.

How does mixedbread help you leverage AI and neural search?

At mixedbread, we offer a suite of services focused on retrieval and embeddings through our simple-to-use APIs. You can use our models to generate high-quality embeddings for a variety of practical use cases and integrate reranking into your search stack to immediately boost performance. We also help solve enterprise needs with customized models, solutions, and local deployments, ensuring your data remains secure and your business stays at the forefront of AI technology. For personalized enterprise solutions, please contact us.

What can our docs help you with?

Our docs section offers a range of resources to deepen your knowledge about embeddings, RAG, and associated topics, as well as the opportunity to learn how to bring the power of our models to your own use case. The quick start section provides everything you need to dive right into our API and get going. You can learn about embeddings and RAG, as well as our flagship embedding model family, in the mixedbread embed section. Enter the realm of reranking in the mixedbread rerank section.