Wikidata JSON Dumps
Wikimedia Foundation
124 million things, from people to proteins, described as public-domain structured facts.
A crowd-edited knowledge graph in which each item (a person, place, work, species, or concept) carries statements such as birth date, country, or occupation, with references and identifiers for hundreds of other databases. The weekly JSON dump holds every item; RDF dumps support SPARQL-style work. It is the usual glue for linking data sets that name the same things.
Access
- Cost
- FreeFree dumps, a public SPARQL query service, and an API.
- License
- Public domainCC0 for structured data in the main, Property, Lexeme, and EntitySchema namespaces
- Sign-up
- No sign-up
- How you get it
- API and Bulk downloadJSON, RDF Turtle, N-Triples, XML
What's in it
- One row is
- One record per item (Q-number) with its labels, statements, and sitelinks
- Covers
- Global
- Years
- 2012-present
- Updated
- Weekly
- Size
- About 124 million items; a full JSON dump was about 130 GB compressed in 2024
Quality
Huge, CC0, and well tooled, but completeness depends on volunteer and bot effort, so coverage is uneven by topic and region.
- Missing statements do not mean a fact is false
- Coverage skews toward Western, notable, and well-documented subjects
- Full dumps are large enough that some decompressors fail on them
Connects to
- Wikipedia Database Dumps: Sitelinks map each item to its Wikipedia articles in every language
- MusicBrainz Database: Wikidata stores MusicBrainz IDs for artists, releases, and works
- WALS (World Atlas of Language Structures): Wikidata language items carry WALS codes for joining typology data
Questions it could answer
- What share of people with Wikidata entries are women, by birth decade and occupation?
- Which countries' museums have the most artworks described in Wikidata?
- How many items link out to each external database?