PT GoTo Gojek Tokopedia Tbk (GoTo or GoTo Group) and Indosat Ooredoo Hutchison (Indosat or IOH) proudly present the latest breakthrough in the Indonesian Large Language Model (LLM) open-source ecosystem, through the launch of Sahabat-AI with a 70 billion parameter capacity model equipped with multilingual chat services.
The introduction of this breakthrough at the Indosat Indonesia AI Day forum in November 2024 marked a significant milestone in strengthening Indonesia’s digital sovereignty, in line with the Indonesian government’s commitment. With an increased capacity of 70 billion parameters, Sahabat-AI now offers higher accuracy, enabling the launch of new chat services accessible through the sahabat-ai.com website and the GoPay application under the “Favorite Citizen Services” menu. The service is equipped with advanced reasoning capabilities, making it easier for users to ask questions and receive informative answers in natural language. This service is just one of the many features and applications that can be utilized from this latest model.
Patrick Walujo, CEO of GoTo Group, stated, “With a 70 billion parameter model and new chat services, Sahabat-AI further strengthens the AI ecosystem that aligns with Indonesia’s characteristics. Its multilingual capability and higher accuracy make this service able to meet the needs of the people and businesses throughout Indonesia. This reflects our commitment to digital sovereignty and supports President Prabowo’s vision for the development of local technology.”
“By developing this model together with Indosat and other partners, we are creating a smarter, faster, and more affordable platform. We have also introduced the Sahabat-AI chat service in the GoPay application used by millions of people, allowing communities across the country to easily benefit from this unique Indonesian LLM. Sahabat-AI has provided significant impact for customers within the GoTo ecosystem by reducing costs, improving service quality, and enhancing interactions. More than just a technological breakthrough, Sahabat-AI has strategic advantages that can drive the progress of Indonesia’s digital economy.”
Vikram Sinha, President Director and CEO of Indosat Ooredoo Hutchison, stated, “Indosat proudly leads the development of AI based on Indonesia’s sovereignty. As part of this initiative, we have introduced GPU Merdeka, an AI cloud that builds a solid digital foundation to ensure AI innovation that is nationally secure, culturally relevant, and accessible to all segments of society. Sahabat-AI is not just a model, it is a national asset supported by collaboration and built for the entire Indonesian people.”
Indosat spearheads the Sahabat-AI ecosystem as a manifestation of its commitment to strengthen and empower Indonesia in the digital era. Starting with the implementation of GPU Merdeka from Lintasarta – Indosat’s AI Factory, to providing real-time multilingual services, Indosat provides the technical infrastructure that serves as the main foundation for training, data processing, and large-scale development of Sahabat-AI. This infrastructure is operated locally, ensuring data remains in Indonesia, maintaining optimal performance, and complying with applicable regulations, which is a crucial factor in supporting the adoption of this technology by the public sector and innovative companies.
The latest Sahabat-AI model can now be used in Indonesian and four regional languages: Javanese, Sundanese, Balinese, Batak, as well as several other international languages. This model can also run optimally on locally accessible infrastructure, allowing various users, from early-stage startup companies, university labs to large-scale public service institutions, to integrate AI into their daily work.
The development of Sahabat-AI is evidence of Indonesia’s spirit of mutual cooperation, uniting research institutions, universities, media organizations, government agencies, and other partners in a national effort. The current collaboration continues with leading universities such as the University of Indonesia, Gadjah Mada University, Bandung Institute of Technology, IPB University, Udayana University, and North Sumatra University, as well as media organizations such as Kompas Group, Republika, Tempo, and Hukumonline, to ensure Sahabat-AI is optimized according to local context and cultural relevance.
Strengthening Indonesia’s digital sovereignty
From the outset, Sahabat-AI has been designed in line with Indonesia’s digital sovereignty goals. All data and GPU infrastructure used to serve this model are stored in Indonesia or on users’ own servers, ensuring compliance with national data regulations. By storing and processing data directly in Indonesia, Sahabat-AI opens up new opportunities for the Indonesian government and public institutions to build secure and sovereign AI services.
Luhut Binsar Pandjaitan, Chairman of the National Economic Council, said, “Data sovereignty is not just a technical issue, but a national independence issue in the digital era. I highly appreciate the initiative of GoTo and Indosat in pioneering Sahabat-AI, and driving technological innovation rooted in our national identity. By developing AI solutions that understand and serve our diverse languages and unique cultures, we take significant steps to ensure that digital transformation benefits all the people of Indonesia.”
Empowering AI technology talents
The development of national talent is at the core of Sahabat-AI’s mission. Therefore, a structured internship program has been launched to help local university students gain direct experience in AI development. Together with technical experts from GoTo and Indosat, the internship participants directly contribute to the training and development of models, acquiring practical skills that are rarely obtained during their educational period. This approach makes Sahabat-AI not only a national innovation but also a platform to advance the next generation of Indonesian AI talent.
One of the internship participants, Komang Ayu from Udayana University, said: “This experience deepened my understanding of end-to-end LLM model development. I learned how to collect and process datasets, explore model architectures, and gained practical insights into how data is prepared and used to train AI models.”
Since the launch of the 8 and 9 billion parameter capacity models, Sahabat-AI has been downloaded over 35,000 times on Hugging Face. For technology developers, LLM can be accessed for free on the Sahabat-AI website and the official Sahabat-AI page on Hugging Face. This open access empowers a broader Indonesian AI ecosystem to build, experiment, and collaborate, thereby fostering the creation of various AI applications tailored to local needs.
Download the Sahabat-AI LLM via the following link: Hugging Face. (*)