Wikidata:SPARQL query service/Alternative endpoints
The Wikidata Query Service at query.wikidata.org is the main endpoint to run SPARQL queries against Wikidata for the general public. There are alternative endpoints that other organizations made available and that you can use as well.
List of available endpoints
[edit]These alternative endpoints are running different software and it is especially useful to better understand which software is working well for which types of queries. Exploring these alternatives is part of the effort to stabilize the Wikidata Query Service and potentially migrate it to a different underlying software in the future. Any support in helping these alternatives grow and improve through testing and development efforts is greatly appreciated.
| Wikidata Query Service | Orb Open Graph | QLever | Virtuoso | MillenniumDB[1] | |
| URL of service | https://query.wikidata.org | https://try.orbopengraph.com | https://qlever.cs.uni-freiburg.de/wikidata | https://wikidata.demo.openlinksw.com/sparql | https://wikidata.imfd.cl |
| Timeout | 60 seconds | None | 600 seconds (?) | 30 seconds | 60 seconds |
| Underlying software | Blazegraph | Blazegraph | QLever | Virtuoso | MillenniumDB |
| Update rate / version | Latest edits incorporated in almost real time | Latest edits incorporated in almost real time | weekly, check date of current version | Dump from 26.04.2024, check | Dump from 17.07.2023, check |
| Graph split? | Separate graphs for publication metadata and everything else | Unified graph | Unified graph | Unified graph | Unified graph |
| Information page with usage policy, privacy policy, issue reporting and other information about the service | Wikidata:SPARQL query service/Wikidata Query Help | Wikidata:Orb Open Graph | Usage and privacy information for the QLever SPARQL endpoints | TODO | Usage and privacy information for the MilleniumDB SPARQL endpoint |
| Deviations from SPARQL 1.1 standard | Complete implementation with custom extensions | Complete implementation with custom extensions | Almost complete, list of deviations and workarounds | All features implemented, various subtle deviations from standard | Core implemented, many features still missing |
Performance evaluation
[edit]The following table summarizes the results of a performance evaluation of the four endpoints above, carried out on 08.04.2024 on 298 example queries from the Wikidata Query Service. For each endpoint, the queries were launched one after the other and the full result was downloaded. For verification, this was repeated twice after a period of a few hours: the results did not deviate significantly in those verification runs.
Columns 2, 3, 4 say which percentage of the queries was processed how fast. Column 5 says which percentage of the queries failed (either due to a timeout or because the endpoint could not process the query for any other reason). Column 6 provides the percentage of queries that did not fail (in parentheses) and their average query time. Note that this gives an undue advantage to those endpoints where many queries fail (it's usually the harder queries that fail), therefore this average should be taken with a grain of salt.
Click here for more results (for other datasets and more SPARQL engines).
| SPARQL endpoint | query time <= 1.0s | (1.0s, 5.0s] | > 5.0s | failed | avg. query time | median query time |
| Wikidata Query Service | 36% of all queries | 20% | 23% | 21% | 6.98s (of 79%) | 2.47s |
| QLever | 78% of all queries | 11% | 9% | 2% | 1.38s (of 98%) | 0.24s |
| Virtuoso | 54% of all queries | 15% | 20% | 11% | 4.11s (of 89%) | 0.74s |
| MilleniumDB[1] | 12% of all queries | 22% | 11% | 55% | 6.05s (of 45%) | > 50% failed |