One row per language carrying an ISO 639-3 code in Wikidata: 8,057 human languages, rostered by the registry key the way US Places used FIPS, so no class-tree guesswork decides what counts as a language. The join keys are the point: ISO 639-3 on every row and Glottolog on 7,918, so the entire linguistics literature attaches cleanly. Around the keys: speaker counts from the latest dated statement (1,929 rows carry one, with the year alongside because freshness matters and because dead languages legitimately carry their last census: Dalmatian's says 1898, the year its final speaker died), scripts, the immediate family branch, how many countries list the language, the region it is indigenous to, and both endangerment scales where recorded (Ethnologue and UNESCO, 93% of rows carry at least one), distilled into an Endangered flag. The headline the flag carries: 44% of assessed languages are at risk, which is the famous half-of-all-languages claim showing up in the graph unprompted. Speaker coverage is honestly thin below the big languages and NA is NA; future-dated speaker counts are typos and lose their dates but not their counts. Built with a 9-test acceptance suite anchored on English's billion-plus, Mandarin, Basque the isolate at 750,000, and the at-risk share landing between 35 and 60 percent.
8,057 rows in one CSV file, with 14 columns in total: Language, Language_ID, ISO_639_3, Glottolog, Family_Branch, Speakers, Speakers_Year, Scripts, Countries_Count, Indigenous_To, Ethnologue_Status, UNESCO_Status, Endangered, Language_URL.
It is built by programmatically scraping wikidata.org, last pulled on 2026-09-22. Datasets are versioned; older versions stay downloadable.
Yes. Every CodeSights dataset is completely free as a CSV download; a free account is all it takes. Anyone can preview the data without signing in.
Yes. The exact Python scraper that built it is viewable on the dataset page by any signed-in member, so every number is reproducible.
Automated scraping leaves room for error and the underlying sources change over time, so no version is guaranteed accurate or complete. If a number matters, verify it against the original source.