GoTo, described as the largest digital ecosystem in Indonesia, and service provider Indosat Ooredoo Hutchison (IOH) have announced the launch of a new, enhanced version of Sahabat-AI, an Indonesian large language model launched in mid-November last year.
Launch versions of Sahabat-AI offered a mere eight or nine billion parameters. Now upgraded to 70 billion parameters, Sahabat-AI offers enhanced accuracy, enabling the launch of the new Sahabat-AI chat service, which is built with strong reasoning capabilities, enabling users to ask questions and receive insightful, natural-language responses. It is one of the many use cases made possible by the more powerful model.
More importantly perhaps, the latest model can now operate across five local languages: national language Bahasa Indonesia, as well as Javanese, Sundanese, Balinese and Bataknese, in addition to a number of international languages.
As news service Forbes points out, this development is part of a drive to preserve local culture and dialects amid a global surge in AI development, currently led by models trained in the US and China.
Of course the term ‘local languages’ may imply modest numbers of speakers. However, local languages in Indonesia can boast tens of millions of speakers. This LLM therefore gives the new service access to many rural Indonesians that may not speak English or even Bahasa Indonesia.
There is, inevitably, a digital sovereignty angle to all of this. All data and GPU infrastructure used to serve the model are stored within Indonesian territory or on users’ own servers, ensuring compliance with national data regulations. By localising data storage and processing, its creators argue, Sahabat-AI also opens up new opportunities for the Indonesian government and public sector to build secure, sovereign AI-powered services.
Indeed, the model is optimised to run on locally accessible infrastructure, enabling a broader range of users, from early-stage startups and university labs to large-scale public service institutions, to integrate AI into their workflows.
For developers, the LLM can be freely downloaded, an open access approach that GoTo and IOH say empowers Indonesia’s broader AI ecosystem to build, experiment, and collaborate – encouraging the creation of AI applications tailored to local needs.
