PT GoTo Gojek Tokopedia Tbk (GOTO), Indonesia’s largest digital ecosystem, has just announced the launch of Sahabat-AI, a groundbreaking open-source Large Language Model (LLM) tailored to Bahasa Indonesia and local languages across the country. The initiative, developed in partnership with PT Indosat Ooredoo Hutchison Tbk, is part of GoTo’s ongoing commitment to advancing technology and digital sovereignty in Indonesia.

Large Language Models, or LLMs, are sophisticated AI systems trained on vast datasets to understand and generate human language. They can be used in a range of applications, from customer service chatbots to more complex data analytics tools. However, most globally available LLMs tend to focus on major world languages like English, often overlooking the unique linguistic nuances and cultural contexts found in other languages, especially in countries like Indonesia.

Sahabat-AI, however, stands out due to its development specifically for Indonesia’s needs. It is not only designed to understand and process Bahasa Indonesia, but it also includes regional languages like Javanese and Sundanese, with plans to extend to languages such as Batak and Balinese. This is a significant leap for the country, given its cultural and linguistic diversity.

Patrick Walujo, the CEO of GoTo, emphasized the importance of localizing AI for Indonesia’s unique challenges. “As a company deeply rooted in Indonesia, we aim to support the country’s growth by providing technology that resonates with the local context. Our partnership with Indosat on Sahabat-AI will fill the gap left by foreign AI models, providing a more culturally relevant and practical tool for local developers, businesses, and communities,” he explained.

Catherine Hindra Sutjahyo, the President of GoTo’s On-Demand Services, elaborated further during the launch at the Indonesia AI Day in Jakarta. She noted, “With Indonesia’s vast population and diverse culture, it is crucial that we have an AI system that truly understands our language and context. Sahabat-AI is built to support this need and will continue to evolve to meet the demands of our local communities.”

One of the key aspects that makes Sahabat-AI different from existing global models is its open-source nature. This means that it is freely accessible for developers, engineers, and even businesses to build custom AI applications suited to Indonesia’s needs, from customer service tools to research platforms. This initiative is seen as an effort to reduce the country’s reliance on foreign AI models while fostering innovation from within. In addition, it is expected to contribute to the preservation and enrichment of local languages, something that many global AI models have overlooked.

In its initial phase, Sahabat-AI is launched with 8 and 9 billion parameters, and early tests show it already outperforms some open-source models in Bahasa Indonesia, Javanese, and Sundanese. Despite still being in the early stages, the project is already demonstrating great potential, and GoTo is inviting local stakeholders, including businesses, universities, and government agencies, to collaborate in further developing this platform.

The Sahabat-AI model can be downloaded for free via Hugging Face, a platform where machine learning communities collaborate on AI projects. Additionally, GoTo has partnered with various Indonesian universities, such as Gadjah Mada University, to engage students in the development of the AI model. These students will not only help with data collection and validation but will also gain valuable experience and training in the field of AI.

Sahabat-AI has already been integrated into GoTo’s existing technologies, such as Dikte Suara (Dira), a voice recognition tool for GoPay, part of GoTo’s fintech ecosystem. Dira allows users to navigate GoPay features and complete tasks faster by simply speaking in Bahasa Indonesia. Looking ahead, GoTo plans to further integrate this technology across its platforms, making it a central part of the user experience.