GigaOm ranks Vespa.ai top in vector database report

Research house GigaOm has assessed 17 vector databases in a Radar report, rating Vespa.ai as the top offering.

A vector database stores and makes accessible vector embeddings, the mathematical representation of aspects or parts of digital text, images, sounds, and videos. The vectors are used in semantic search by AI large language models (LLMs). Broadly speaking, these take in a natural language search term, vectorize it, and then search for similar vectors in the databases, from which they generate their response.

GigaOm’s 17 vector database suppliers were Activeloop, AWS, Chroma, Google, IBM, LanceDB, Marqo, Microsoft, MongoDB, OpenSearch, Oracle, Pinecone, PostgreSQL, Qdrant, Vespa.ai, Weaviate, and Zilliz. SingleStore was not included in the list, although it has vector store and retrieval facilities in its database.

GigaOm’s Radar diagram plots vendor offerings across a series of concentric rings – Entrant, Challenger, and Leader – with those placed nearer the center rated as being most complete. Products are positioned in two axes, balancing Maturity vs Innovation on one and Feature Play vs Platform Play on the other. An arrowhead symbol projects a product’s progress over the next 12 to 18 months, with three classifications: Forward Mover, Fast Mover, and Outperformer.

The majority, 14, are placed in the innovation area, as they are still rapidly developing and not mature offerings, with nine in the platform area. The Leaders are Vespa.ai, which was ahead of IBM, Zilliz, Qdrant, Weaviate, OpenSearch, and MongoDB.

Several suppliers did not respond directly to the two GigaOm analysts, Andrew Brust and Jelani Harper, and their evaluation was carried out using desk – document and website – research. Those vendors were Google, Marqo, Oracle, Pinecone, and PostgreSQL.

There are two general types of vector database suppliers. Startups offer dedicated vector database facilities, such as Pinecone, Qdrant, Vespa.ai, Weaviate, and Zilliz. They will stress that their database structures and functions are expressly focused on vector search and retrieval, and provide speed and developer ease of use. The second type consists of existing database suppliers that add vector storage and retrieval to their offering, and will talk about content type integration inside their database, simpler overall database admin, and facilities for transforming their existing data into vectors. Examples include IBM and SingleStore.

Public cloud suppliers Amazon, Google, and Microsoft (Azure) will also stress the integration of their vector store and search capabilities with their existing offerings.

Vespa.ai has made the GigaOm Vector Database Radar report available for download here.