At Bali Tech Week on October 19, GoTo released a 27-billion-parameter large language model trained for Bahasa Indonesia and regional languages.

It described the model as open source and freely available to use and adapt.

GoTo, formally PT GoTo Gojek Tokopedia Tbk, was formed through the merger of Gojek and Tokopedia.

The company said its training data came from media partners and public-sector websites, with Universitas Sumatera Utara helping gather Batak-language data.

The release builds on GoTo’s open-model work that began in 2024, when it launched Sahabat-AI with Indosat Ooredoo Hutchison as part of a digital sovereignty push.

GoTo said the model can run on institutions’ own servers.

That may appeal to public bodies in Indonesia, where analyses of the country’s data-localization rules say some public-sector and strategic electronic-system data must be stored and processed within Indonesia, though that does not necessarily require on-premise hosting.

🔗 Source: GoTo

Recent GoTo developments