How many languages does Google Translate support?

How to get close to this number

Think about which languages have enough digital text to train on. A machine translation system historically needed large parallel corpora — the same text in two languages — which limited it to languages with substantial published translation. That constraint put the count in the low hundreds for years.

The explanation names what broke that ceiling: over a hundred languages added in a single update, using models trained without parallel texts. When the data requirement changed, the set of feasible languages changed with it, and a figure that had grown slowly for a decade jumped in one step. Technology counts often move in steps rather than curves, and the step marks a method change.

The coverage claim is the useful framing: languages spoken by roughly ninety-five percent of the world population. That sounds close to complete and is not, because the remaining five percent is spread across thousands of languages — exactly the skewed distribution the languages question on this site describes. Population coverage and language coverage are very different measures of the same system.