I’ve been hearing a bit about Weaviate, and it has me wondering whether anyone has seen a comparison chart of the vector database options. These databases store the knowledge that custom-purpose AI chatbots use — the part that makes a bot yours rather than everyone’s.
Some context, because the money noticed this category before most builders did. Weaviate, which is open source, raised a $50 million Series B in April, led by Index Ventures, and the company says it “solves the hard problem of generating, storing, and searching embedding vectors.” Six days later Pinecone raised $100 million at a $750 million valuation, led by Andreessen Horowitz. Pinecone’s chief executive put the relationship plainly: “think of the LLM as an application almost that sits on top of this database.”
That is the whole idea in one sentence. You turn your documents into vectors — long lists of numbers that stand for meaning — and store them. When a customer asks something, the database finds the passages nearest in meaning, and those go to the model along with the question.
Also, on Weaviate: one place to start is ANN-Benchmarks, which describes itself as “a benchmarking environment for approximate nearest neighbor algorithms search.” It pits the search algorithms underneath these databases against one another on speed and accuracy. It measures search speed and accuracy, not the whole product — nothing about pricing, hosting or how the dashboard feels at two in the morning — so it’s a rigorous place to begin and a poor place to stop.
The chart that’s missing would cover the rest: cost, hosting, and how each of these behaves once you’ve loaded a few million vectors and real customers start asking questions. Has anyone seen one?
