What Is a Vector Database?

The short definition

A vector database stores data as embeddings — long lists of numbers that represent the meaning of a piece of text, an image, or other content — and it's built to search those embeddings by similarity rather than by exact match. Ask it for "things similar to this," and it returns the nearest neighbors in that numeric space, which is what lets it find conceptually related content even when the wording is completely different from your query.

Pinecone and Qdrant are two common implementations of this idea, and they take opposite approaches to how you run one. Pinecone is a fully managed service priced on storage plus read and write units. Qdrant is open source, so you can self-host it for free, and it also offers a managed cloud tier that starts free and moves to usage-based pricing above that.

How it actually works

Under the hood, a vector database indexes embeddings so that finding the closest matches to a query doesn't require comparing against every stored vector one by one — that's what makes similarity search fast even across millions of entries. You write data in as embeddings, and you read it back out by submitting a query embedding and getting the nearest matches.

Pinecone bills each of those operations separately: writes and reads are metered in units, and storage is billed separately from both. The free Starter plan includes 2 GB of storage, 2M write units, and 1M read units a month across up to 5 indexes, which is enough to build and demo a real application before paying anything. Above that, Builder is a flat $20/mo for up to 10 GB storage, 5M writes, and 2M reads a month, and Standard moves to metered pricing with a $50/mo minimum spend, storage at $0.33 per GB per month, and reads costing $16 to $18 per million units — the most expensive dimension for a read-heavy application.

Qdrant's cloud free tier works differently: it's free forever rather than time-limited, giving you 0.5 vCPU, 1 GB RAM, and 4 GB disk on a single-node cluster, plus free cloud inference with selected models. Beyond free, Standard and Premium tiers are usage-based without a published fixed price — Standard adds backup, disaster recovery, and a 99.5% uptime SLA, while Premium adds SSO, private VPC links, and a 99.9% uptime SLA. Because Qdrant is open source, you can also skip the cloud product entirely and self-host it yourself for free.

When you need it — and when you do not

You need a vector database when you're building semantic search, recommendations, or a RAG pipeline where retrieving conceptually similar content matters more than exact keyword matching. It's the standard retrieval layer for grounding an LLM in your own documents — see our RAG explainer for how that pairing works end to end.

You don't need one for small datasets that fit comfortably in memory or for search that's genuinely keyword-based, where a traditional database index does the job without the added infrastructure. Scale and read patterns also matter for which vendor fits: Pinecone's free tier's 1M monthly reads suits a real prototype, but read-heavy production traffic hits Pinecone's $16-to-$18-per-million read cost fastest, so it's worth estimating query volume before committing to a managed, metered service.

Qdrant's free cloud tier is capped at 1 GB RAM and a single node, which is genuinely not enough for a production workload — it's a prototyping tier, not a production one. If self-hosting appeals because you want to avoid a vendor relationship or keep data fully on your own infrastructure, Qdrant's open-source edition and Hybrid Cloud option (priced on request) are built for exactly that, at the cost of owning your own sizing, upgrades, and backups.

The tools that do this

Pinecone and Qdrant are the two vector databases with verified pricing here, and the practical choice mostly comes down to whether you want fully managed or the option to self-host. Pinecone's tiered, metered pricing (free Starter, flat $20/mo Builder, metered Standard from a $50/mo minimum, and Enterprise from a $500/mo minimum) gives you a predictable ladder as usage grows, with no cluster to size or patch yourself.

Qdrant's model is more flexible on cost at the extremes: free forever on its cloud tier for small workloads, completely free if you self-host the open-source edition, but usage-based and unpublished for its Standard and Premium cloud tiers, so you'll need its pricing calculator to get an exact number once you're past the free tier. If you're prototyping a RAG application, both free tiers are enough to start; the decision to switch usually comes down to whether you want a managed service to handle scaling for you, or you're comfortable owning the infrastructure yourself in exchange for lower cost at scale.

There's also a middle path worth knowing about: Qdrant's Hybrid Cloud option lets you run managed clusters on your own infrastructure, which is a different tradeoff than either fully managed cloud or fully self-hosted. It's priced on request rather than published, so it's a conversation with the vendor rather than a number you can plan around today, but it's a reasonable option for a team that wants Qdrant's operational tooling without handing data off to a third-party cluster.

The bottom line for picking between them isn't really about which vendor is "better" in the abstract — it's about how much operational ownership you want. A team that would rather pay for reads and writes and never think about cluster health again fits Pinecone's model well. A team that wants to avoid vendor lock-in, or that already runs its own infrastructure and doesn't mind adding one more service to it, gets more value out of Qdrant's open-source option. Either way, start on the free tier, run a real workload through it, and let actual read and write volume — not the pricing page alone — tell you which one scales more cheaply for your specific use case.