Try asking a popular AI chatbot something in Javanese or Balinese, and you’ll quickly notice the difference. Most large AI models are trained mostly on English text, so they often miss local expressions, culture, and context. For a country with hundreds of regional languages, that’s a real problem.

That’s why Sahabat-AI exists. Developed in Indonesia by Indosat Ooredoo Hutchison and GoTo, Sahabat-AI is a family of open-source large language models built for Bahasa Indonesia and several regional languages. In 2026, it’s still at the center of Indonesia’s “sovereign AI” conversation. Here’s what it is, how it works, and why it matters for the next generation of Indonesian engineers.

What’s Happening With Sahabat-AI

Sahabat-AI was launched on November 14, 2024, at the Indonesia AI Day conference in Jakarta. It was built by Indosat and GoTo with support from Nvidia and several other partners. The first models were open source, meaning developers and companies can download them, adapt them, and build their own apps on top.

The project has grown since then. In June 2025, Fortune reported a new 70-billion-parameter version that supports Bahasa Indonesia, Javanese, Balinese, and Batak, among other local languages. The models run on GPU Merdeka, a sovereign AI cloud from Indosat’s subsidiary Lintasarta, powered by Nvidia GPUs, so data can stay inside Indonesia.

In a May 2026 interview with Fortune, Indosat CEO Vikram Sinha explained why a local model matters: “The LLM is not neutral. And if it is not in your language, it will have bias, cultural nuances and all.” He also admitted the business model is still being worked out, describing Sahabat-AI as something closer to a public service for now.

The Tech Behind It: How a Local AI Is Built

A large language model (LLM) is a program that learns patterns in language by studying huge amounts of text. Its “parameters” are the internal numbers it adjusts during training, and more parameters usually means it can capture more complex patterns. Sahabat-AI’s first version started from Meta’s open Llama 3 model with 8 billion parameters and was then fine-tuned, or further trained, on Indonesian and regional-language data.

Fine-tuning is a smart shortcut. Instead of building a model from zero, which costs enormous amounts of computing power, engineers take an existing model and teach it new skills with carefully chosen data. The team used Nvidia’s NeMo tools to customize the model and Nvidia NIM software to run it more efficiently.

The infrastructure matters just as much as the model. Training and running LLMs needs racks of GPUs in a data center, fast networking, and reliable power. By hosting Sahabat-AI on GPU Merdeka in Indonesia, the project keeps sensitive data within the country, an idea known as “sovereign AI.”

What It Means for Future Engineers

Sahabat-AI shows that Indonesia doesn’t have to just import AI. It can build and adapt AI for its own people. Banks, government services, startups, and schools can all benefit from AI that understands local languages and culture.

But building something like this takes a wide range of skills. Someone has to prepare and clean the language data, someone has to train and test the models, and someone has to design and run the GPU servers, networks, and cooling systems underneath. Then there are the engineers who bring AI to phones, smart devices, and embedded systems, where memory and power are limited.

That combination of software, hardware, and systems thinking is exactly what computer engineering is about. As more Indonesian companies build their own AI, they’ll need engineers who understand the whole stack, from the chip all the way up to the app.

Key Takeaways

Sahabat-AI, launched in November 2024 by Indosat and GoTo, is an open-source family of AI models built for Bahasa Indonesia and regional languages such as Javanese, Balinese, and Batak. It grew from an 8-billion-parameter model to a 70-billion-parameter version and runs on sovereign AI infrastructure inside Indonesia. Its story shows that local AI needs local engineers, from data and models to the hardware they run on.

Want to help build AI that truly understands Indonesia? The Computer Engineering program at BINUS University prepares you to work across hardware, software, and intelligent systems. Explore the program and be part of Indonesia’s AI future.

Sources