Indosat & GoTo launch Sahabat-AI LLM for five languages

Indosat Ooredoo Hutchison and GoTo Group have launched Sahabat-AI's 70-billion parameter large language model alongside a multilingual chat service.

The companies announced that the Sahabat-AI model is now capable of operating across five local Indonesian languages: Bahasa Indonesia, Javanese, Sundanese, Balinese, and Bataknese, aiming to cater to millions of users throughout Indonesia.

This development builds on Sahabat-AI's initial release in November 2024 and reflects ongoing efforts to bolster Indonesia's digital sovereignty in alignment with national government objectives. According to Indosat and GoTo Group, the enhanced model offers greater accuracy and reasoning capabilities, supporting a range of use cases, including the launch of a chat service available on sahabat-ai.com and the GoPay app.

Patrick Walujo, Chief Executive Officer of GoTo Group, said, "With the launch of our 70 billion parameter model and the new chat service, Sahabat-AI takes a major leap forward in building a uniquely Indonesian AI ecosystem. Its multilingual capability, combined with enhanced accuracy, enables Sahabat-AI to better serve the diverse needs of people and businesses across the country. It reflects our commitment to digital sovereignty and supports President Prabowo's vision for locally developed, locally hosted technology."

"By building this model in partnership with Indosat and a broad range of partners, we've created a platform that is smarter, faster, and more affordable. We're also making the Sahabat-AI chat service available through the GoPay app, which is used by millions, ensuring people across the country can easily benefit from this uniquely Indonesian LLM."

"Sahabat-AI is already making a significant impact on customers across the GoTo ecosystem, lowering costs, improving service quality, and deepening engagement. More than a technological breakthrough, Sahabat-AI is a strategic advantage for us, and a powerful engine of growth for Indonesia's digital economy."

The model is optimised to function on locally accessible infrastructure. This approach enables broad adoption, allowing early-stage startups, university research labs, and public sector institutions to integrate AI into their workflows.

Vikram Sinha, President Director and Chief Executive Officer of Indosat Ooredoo Hutchison, said, "Indosat is proud to lead the development of Indonesia's sovereign AI capabilities. Through GPU Merdeka, our sovereign AI cloud, we're laying the digital foundation to ensure that AI innovation is not only advanced but also nationally secured, culturally relevant, and equitably accessible. Sahabat-AI is more than a model — it's a national asset powered by collaboration and built for all Indonesians."

Indosat's involvement includes establishing the Sahabat-AI ecosystem, deploying the GPU Merdeka platform, and providing support for real-time multilingual services. The underlying infrastructure is locally operated to ensure data residency, performance, and regulatory compliance, providing a pathway for public sector adoption and wider enterprise usage.

The model operates in five local languages as well as several international languages. Its efficiency allows operation on two H100 GPUs, compared to larger open-source models that often require more than sixteen GPUs, which broadens accessibility through reduced resource requirements.

The collaboration with Indonesian research institutions and universities is an integral part of Sahabat-AI's development. Partners include University of Indonesia, Gadjah Mada University, Bandung Institute of Technology, IPB University, University of Udayana, and University of Sumatera Utara. Media partners, including Kompas Group, Republika, Tempo, and Hukumonline, are also involved to ensure local context and cultural relevance in the model's outputs.

The companies emphasised the importance of national digital sovereignty for Indonesia. Sahabat-AI's data and GPU infrastructure are maintained within Indonesian territory or on users' own servers. This setup ensures alignment with national data regulations and provides additional opportunities for secure, AI-powered public sector services.

Luhut Binsar Pandjaitan, Chairman of the National Economic Council, said, "Data sovereignty isn't just a technical matter, it's a matter of national independence in the digital era. I applaud GoTo and Indosat for pioneering the ecosystem of Sahabat-AI, and fostering technological innovation rooted in our national identity. By developing AI solutions that understand and cater to our unique yet diverse linguistic and cultural landscape, we are taking a significant step towards ensuring digital transformation benefits all Indonesians."

A central element of the Sahabat-AI initiative is talent development through a structured internship programme. Local university students work directly with engineers from GoTo and Indosat, gaining practical experience in AI development. According to the companies, this addresses the gap between classroom learning and industry requirements for large language model development.

Komang Ayu from the University of Udayana, who participated as an intern, said, "This experience deepened my understanding of the end-to-end development of large language models. I learned how to collect and preprocess datasets, explored model architectures, and gained practical insights into how data is prepared and used to train AI models."

Since earlier releases of the Sahabat-AI model with 8 and 9 billion parameters, the models have been downloaded more than 35,000 times on the AI repository platform Hugging Face, the companies reported. The models are available for download both from the Sahabat-AI website and the official Hugging Face page, which aims to provide open access and encourage experimentation and customised application development across Indonesia's AI community.