The right replacement depends on whether you want to run the database yourself and how big your data will get. Open-source vector engines suit teams that self-host, search engines with vector support suit app search plus retrieval, and AI data platforms suit large training sets. Test storage, query speed and filtering on your own data first.
Do you want Milvus as a managed service, or do you want a vector database that fits your stack in a different way? That's the real question behind most searches for Zilliz alternatives. Here are seven vector databases and AI search engines for developers storing and searching embeddings, in no particular order, with notes on how each is run and what it's built for.
Managed lakebase or a database you run yourself
Zilliz's own homepage describes a fully managed vector lakebase powered by Milvus, covering real-time vector search, lake-scale discovery and AI data operations across the big three cloud providers. It's a managed service first, and that shapes everything about the decision to leave it.
If you're moving away, you're choosing between two paths. You can run an open-source engine yourself, which gives you control over where data lives and how it's tuned, but hands you the upgrades, backups and scaling. Or you can use another managed service, which takes that work away but ties you to someone else's limits. Several tools below offer both, with an open-source core and a hosted cloud on top.
Whichever path you take, the thing to test is how data is stored and queried at scale. A demo with a few thousand vectors tells you very little. Load a realistic slice of your own embeddings, add the metadata filters your app needs and measure query speed and recall as the collection grows.
Seven vector stores, alphabetically
Listed A to Z and in no particular order. Each line reflects the vendor's own description of its product.
Chroma
Best for: Developers building AI apps who want open-source search infrastructure they can run and inspect themselves; the vendor names a free plan.
LanceDB
Best for: Machine learning teams curating large, multimodal datasets and doing feature engineering, which the vendor says helps teams train better models faster.
Meilisearch
Best for: Product teams wanting one engine for search and AI retrieval, run either fully managed on the vendor's cloud or self-hosted from open-source code.
Qdrant
Best for: Engineers who want an open-source vector search engine written in Rust, with fast, scalable similarity search through what the vendor calls a convenient API.
Typesense
Best for: Developers who need fast, typo-tolerant site and app search with simple APIs, vector search and the option to self-host, per the vendor.
Vespa
Best for: Teams building AI search, agents, personalisation and recommendations on one platform the vendor describes as fast and accurate.
Weaviate
Best for: Builders of AI-native apps who want fewer hallucinations, less data leakage and less vendor lock-in, as the vendor frames it; a free plan is named.
Vector engines, search engines and AI data platforms
Dedicated vector engines
Qdrant, Weaviate and Chroma are built for embeddings first. They're open source at the core and suit developers building retrieval for chatbots, assistants and semantic search who want a database designed around vectors.
Search engines with vectors added
Typesense, Meilisearch and Vespa started as search engines and now handle vector and AI retrieval alongside keyword search. They suit teams that need site or app search and AI retrieval from the same engine, so typo tolerance and filters come with the embeddings.
AI data platforms
LanceDB is aimed at a different job: curating large, multimodal datasets for training and feature engineering. It suits machine learning teams whose main problem is data preparation rather than serving queries to an app.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Seven options in a quick table
| Tool | Kind of tool | Good for |
|---|---|---|
| Chroma | Open-source AI search | Open-source retrieval for AI apps |
| LanceDB | Multimodal AI lakehouse | Curating large training and feature datasets |
| Meilisearch | Search engine, cloud or self-hosted | Site and app search plus AI retrieval |
| Qdrant | Open-source vector search engine | Fast similarity search through a simple API |
| Typesense | Open-source typo-tolerant search | Site and app search with vector support |
| Vespa | AI search platform | Search, recommendations and retrieval at scale |
| Weaviate | AI-native vector database | Complete AI apps with less lock-in |
Advice for developers replacing a vector database
Don't choose on benchmark charts. Every vendor publishes the tests that suit it, and none of them use your data. Take a realistic sample of your own embeddings, your own filters and your own query mix, and run them on two or three shortlisted tools. Measure recall as well as speed, because a fast answer that misses the right document isn't much use.
Think about who'll look after the database, too. If you don't have someone who enjoys running infrastructure, a managed service is the safer start, and most open-source options here offer one. For more background, see my guide to the best vector databases and my roundup of the best RAG tools if retrieval for a chatbot is the goal.
Self-hosted or managed: questions to settle first
Where must the data live?
If customer data can't leave your own cloud account or region, self-hosting or a vendor that deploys into your account narrows the field quickly. Data rules vary by country, so check with your data regulator if personal data is involved.
How big will the collection get?
Estimate your vector count and query volume a year out, not just today. Some engines are happiest on a single machine, while others are built to spread across many.
Do you need keyword search too?
If users type queries with typos or product codes, hybrid search that mixes keywords and vectors will help. Search engines with vector support handle that natively.
Do you make vector database or AI search software? Tool makers can ask to be considered for a featured place on this list.
Want your tool on this list? Featured spots and reviews here are paid partnerships. Here's how to partner with me.
Frequently asked questions
What is a vector database?
It stores embeddings, which are lists of numbers that represent the meaning of text, images or other data, and finds the closest matches to a query. It's the retrieval layer behind most AI chatbots and semantic search.
Is Milvus the same as Zilliz?
Milvus is the open-source vector database, and Zilliz offers a fully managed service built on it. You can also run Milvus yourself if you'd rather self-host.
Should I self-host a vector database?
Self-hosting gives you control over data location and tuning, but you handle upgrades, backups and scaling. If nobody on the team wants that job, start with a managed service.
What is hybrid search?
Hybrid search combines keyword matching with vector similarity, so exact terms like product codes and loose meaning both count. It often gives better results than either alone.
How do I move embeddings to a new vector database?
Export the vectors with their IDs and metadata, then load them into the new store in batches. If you change embedding models at the same time, you'll need to re-embed everything.
Do I need a separate vector database for a small project?
Not always. For a small collection, a search engine with vector support or a vector extension on your existing database may be enough. Move to a dedicated store when size or speed demands it.
