Softwr

Databases · head to head

DuckDB vs Milvus

DuckDB logo

DuckDB

Databases

MIT-licensed analytical SQL database that runs inside your process, with no server, no dependencies and one writer at a time.

From
Free
Rated
-
Full reviewWebsite
Milvus logo

Milvus

Machine Learning

Open-source vector database for scalable similarity search

From
Free
Rated
-
Full reviewWebsite

The short version

  • Each has a real cost: DuckDB a database file accepts one read-write process at a time; other processes must open it read-only and will not see subsequent writes, so DuckDB cannot be the shared database behind several services.; Milvus vector dimensions are capped at 32,768
  • They diverge on capability: DuckDB covers In-process execution, Milvus covers Billion-scale vectors.
  • Prices and features above were last checked on 30 August 2026.

Where they differ

Only the attributes on which DuckDB and Milvus actually diverge.

Attributes where DuckDB and Milvus differ
AttributeDuckDBMilvus
Pricing modelopen-sourcefreemium
PlatformsLinux, macOS, Windows, WebAssemblyLinux, Mac, Windows, Web
CategoryDatabasesMachine Learning
Founded20192017

Identical on both: starting price (Free), free tier (Yes), user rating (Not yet rated).

What each one covers

Drawn from each product's published feature list. An absence here means we hold no record of it - not that the product lacks it.

Only in DuckDB

  • In-process execution
  • Vectorised columnar engine
  • Direct file querying
  • Zero dependencies
  • Larger-than-memory queries
  • MIT licence
  • Postgres-flavoured SQL
  • Extension ecosystem

Only in Milvus

  • Billion-scale vectors
  • Multiple index types
  • GPU acceleration
  • Hybrid search
  • Data partitioning
  • PyTorch
  • TensorFlow
  • Hugging Face

What people use each for

The jobs each tool is most often brought in to do.

DuckDB

  • Transformation steps in a data pipeline that would otherwise need Spark, replaced by SQL over Parquet in a single processnot Milvus
  • Analytical queries embedded in an application or a dashboard where shipping a database server alongside it is not acceptablenot Milvus
  • Local exploration of files that are too large for a pandas dataframe but far too small to justify a warehousenot Milvus
  • Continuous integration and testing of analytical SQL, where a real engine can run in the test process without provisioning anythingnot Milvus

Milvus

  • Self hosting a vector database for semantic searchnot DuckDB
  • Storing and querying embeddings for retrieval augmented generationnot DuckDB
  • Similarity search over images, audio or text at scalenot DuckDB

Where each one falls short

Documented limitations, not opinions. Every one is a constraint you would hit in normal use.

DuckDB

  • A database file accepts one read-write process at a time; other processes must open it read-only and will not see subsequent writes, so DuckDB cannot be the shared database behind several services.
  • There is no network protocol, authentication or user management, so exposing it to remote clients means writing and securing your own service around it.
  • It is built for scans and aggregations, not for many small transactions, so a workload of high-frequency single-row inserts and updates performs badly compared with SQLite or Postgres.
  • Storage files are backwards compatible but not forwards compatible, so a file written by a newer version cannot be read by an older one and every consumer of a shared file must be upgraded together.
  • Query memory settings matter: some operations still need to hold significant state, so an under-configured memory limit turns a large join or a high-cardinality aggregation into a spill-heavy query or an out-of-memory failure rather than a slow success.

Milvus

  • Vector dimensions are capped at 32,768
  • A collection is limited to 64 fields, 1,024 partitions and 16 shards
  • Only 1 index is allowed per field
  • Search returns at most 16,384 vectors as top-k, and nq is capped at 16,384
  • Input and output per RPC is capped at 64 MB for insert, search and query
  • VARCHAR values are limited to 65,535 characters
  • Data loaded into query nodes cannot exceed 90% of available memory
  • An instance supports at most 65,536 collections

Pricing, plan by plan

DuckDB

Free

No published plan breakdown. See the DuckDB review.

Milvus

Free
  • Open SourceFree
    • Full features
    • Self-hosted
    • Community support
  • Zilliz CloudFree
    • Managed service
    • Free tier available

Which should you pick?

Choose DuckDB if

  • You need in-process execution.
  • You want to start without paying.
  • You work on Linux, macOS, Windows, WebAssembly.
  • You also want vectorised columnar engine.
DuckDB review

Choose Milvus if

  • You need billion-scale vectors.
  • You want to start without paying.
  • You work on Linux, Mac, Windows, Web.
  • You also want multiple index types.
Milvus review

Questions people ask

Is DuckDB or Milvus better?
Neither clearly leads. DuckDB starts at Free and Milvus at Free, and user ratings are close enough to be indistinguishable. Choose on capability and platform support.
Which is cheaper, DuckDB or Milvus?
DuckDB starts at Free and Milvus at Free.
Does DuckDB or Milvus run on more platforms?
DuckDB runs on Linux, macOS, Windows, WebAssembly. Milvus runs on Linux, Mac, Windows, Web.
Can I use DuckDB for free?
Both have a free tier, so you can try either at no cost before committing.
What is DuckDB best used for?
DuckDB is most often used for transformation steps in a data pipeline that would otherwise need spark, replaced by sql over parquet in a single process, analytical queries embedded in an application or a dashboard where shipping a database server alongside it is not acceptable, local exploration of files that are too large for a pandas dataframe but far too small to justify a warehouse, continuous integration and testing of analytical sql, where a real engine can run in the test process without provisioning anything. Of those, transformation steps in a data pipeline that would otherwise need spark, replaced by sql over parquet in a single process and analytical queries embedded in an application or a dashboard where shipping a database server alongside it is not acceptable are not what Milvus is typically brought in for.
What can DuckDB do that Milvus cannot?
DuckDB covers In-process execution, Vectorised columnar engine, Direct file querying, Zero dependencies. Milvus covers Billion-scale vectors, Multiple index types, GPU acceleration, Hybrid search.

Answered from the vendors’ own pages

DuckDB: Can multiple applications share one DuckDB database?

Not for writing. One process holds the database read-write; others may attach read-only and will not see later writes. Shared multi-writer access needs a different database or a table format with a catalogue.

Milvus: How much does Milvus cost?

Milvus is open-source and free to use and modify. The self-hosted version has no licensing cost. Zilliz Cloud (the managed SaaS version) does not publish pricing on the website.

Source
DuckDB: Is it a replacement for a data warehouse?

For single-node analytical workloads up to a few hundred gigabytes it very often is. It is not a replacement when many concurrent users need a shared, governed, always-on service.

Milvus: Is there a free or open-source version of Milvus?

Yes, Milvus is fully open-source and available for free. Milvus Lite is a lightweight option for learning and prototyping that can be installed via pip.

Source
DuckDB: Do I have to load data into it?

No. It queries Parquet, CSV, JSON and Arrow in place, including on object storage. Its own storage format is optional and mainly useful when you want indexes, constraints and faster repeated access.

Milvus: Does Milvus offer a managed cloud service?

Yes, Zilliz Cloud is a fully managed Milvus cloud offering with serverless and dedicated cluster options. Pricing must be requested from the company as it is not listed on the public website.

Source
DuckDB: What is MotherDuck's relationship to it?

MotherDuck is a separate company offering a managed and hybrid service built on the DuckDB engine. DuckDB itself remains MIT-licensed and independent of it, with the IP held by the DuckDB Foundation.

DuckDB: Is it suitable for OLTP?

No. It is designed for analytical scans. For transactional workloads with frequent small writes, SQLite or Postgres is the right tool.

Share

Related pages

More on DuckDB

More on Milvus

Keep looking

Other head to heads